Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The Monetary Transmission Mechanism in the Tropics: A Case Study Approach

The Monetary Transmission Mechanism in the Tropics: A Case Study Approach Abstract Many central banks in low-income countries in Sub-Saharan Africa are modernising their monetary policy frameworks. Standard statistical procedures have had limited success in identifying the channels of monetary transmission in such countries. Here we take a case study approach and centre on a significant tightening of monetary policy that took place in 2011 in four members of the East African Community: Kenya, Uganda, Tanzania and Rwanda. We find evidence of the transmission mechanism in most of the countries. Variations across countries can be explained mainly by differences in the policy regime. Skillful exploitation of natural experiments that provide identifying variation in important variables represents the best hope for increasing our empirical understanding of macroeconomic fluctuations. While lacking the scientific pretension of an explicit probability model, careful historical discussions of events surrounding particular monetary changes, such as those provided by Friedman and Schwartz (1963), persuade precisely because they succeed in identifying relevant natural experiments, and describing their consequences. (Summers, 1991, p. 130) 1. Introduction Macroeconomic outcomes in many Low-Income Countries (LICs) in Sub-Saharan Africa (SSA) have greatly improved in recent years. Many central banks in the region are now looking to modernise their frameworks to make policy more forward looking, to further promote macroeconomic stability and ultimately growth and financial development. At the same time, uncertainty about how to proceed abounds. At the heart of this uncertainty is the question of how different LICs are in ways that matter for monetary policy. How does monetary policy transmit to the economy? An influential line of argument proposes that monetary transmission in LIC economies is weak and unreliable, greatly limiting the scope for monetary policy to serve as a key policy tool for macroeconomic stabilisation.1 Indeed, studies on the effects of monetary policy in activity and prices in LICs have largely relied on the use of statistical techniques such as VARs and often find weak or insignificant effects of monetary policy. This may reflect the nature of the underlying transmission mechanism. However, it may also reflect limitations of the VAR and related techniques in the LIC environment.2 The challenges to identifying the transmission mechanism in the data are big anywhere and probably larger in LICs. As elsewhere, policy is endogenous to events in the economy. More than elsewhere, the structure of the economy itself is evolving, for example as the financial system develops. Meanwhile, the transmission of policy itself depends on the policy framework, which is sometimes both obscure and changing. Finally, large supply shocks are frequent and data are noisy and scarce. Thus, analysis based on VARs—which require relatively long times series with consistent policy frameworks and plausible identification strategies—is unusually challenging.3 In this paper, we attempt, in the words of Summers (1991), to learn by ‘identifying relevant natural experiments and describing their consequences.’ In particular, we attempt to learn about the transmission mechanism by looking closely at a set of related cases in which policymakers in four countries (Kenya, Uganda, Tanzania and Rwanda, hereafter the EAC4) suddenly and unexpectedly tightened monetary policy to varying degrees. In two cases, at least, the policy adjustment was drastic. We thus apply a version of what Romer and Romer (1989) call the ‘narrative approach’ to identifying the effects of monetary policy: ‘The central element of this approach is the identification of monetary shocks through non-statistical procedures… The method involves using the historical record… to identify episodes when there were large shifts in monetary policy or in the behaviour of monetary policy that were not driven by developments on the real side of the economy.’4 To the best of our knowledge, this is the first attempt to apply a case-study methodology to monetary policy transmission in low-income countries. Some of the same features of the monetary policy environment in low-income countries that make VAR-based analyses extremely challenging also greatly complicate our approach too, however. In part because of the rapid evolution of monetary policy regimes, we are not able to identify several independent and comparable tightening events. And it is hard to argue, as do Romer and Romer (1989), that the tightening episodes are unrelated to recent economic developments, notably supply shocks.5 Thus, we rely mainly on a close reading of the narrative and the data to attempt to identify an unexpected component to the monetary policy tightening in question. We complement this narrative approach with two empirical exercises designed to help disentangle the independent role of monetary policy. First, we analyse high-frequency data on the supply shocks—to international commodity prices and global risk appetite—that are the main potentially confounding variables. This allows us to disentangle the timing of these supply shocks, changes in monetary policy, and movements in the exchange rate.6 Second, while we cannot statistically identify monetary policy shocks, we can take advantage of the exogeneity of the relevant supply shocks to identify the effects of these supply shocks per se on the exchange rate and inflation, conditional on the absence of monetary policy shocks. These exercises leave a significant role for monetary policy shocks themselves in explaining the dynamics of these variables, supporting our interpretation of the role of monetary policy shocks themselves. We are able to draw on variations across the four countries studies—notably in terms of economic structure and policy regime—to shed light on the influences of these factors on monetary policy transmission. Two ideas emphasised in the recent literature are salient. Montiel et al. (2012) argue that structural features of the LIC environment, notably underdeveloped and monopolistic financial systems and inflexible exchange rates, are likely to make transmission weak and unreliable, because policy rates do not transmit to lending rates in underdeveloped and monopolistic banking systems, and in any event these rates do not matter that much to the real economy. Other papers, e.g., Berg et al. (2015), and International Monetary Fund (2015), emphasise the role of policy regimes themselves in determining the nature of monetary transmission.7 These regimes in LICs tend to be complex and opaque, with a multiplicity of instruments (interest rates, quantities, foreign exchange intervention, etc.) and objectives (inflation, output, credit, exchange rates, etc.), and with the weights on both instruments and objectives often hard to discern and time-varying. This opacity has implications for transmission itself: where monetary aggregates are targeted, interest rate movements may not signal the policy stance clearly, for example.8 It also has implications for the analyst: even if a clear regime can be identified, it is likely to be of too-short duration to permit econometric analysis.9 Our narrative centres on a significant tightening of monetary policy that took place—to varying degrees and in different ways—in October 2011 in the EAC4. In 2010–2011, a major commodity price shock hit, and inflation took off in the EAC4, echoing the events of 2007–2008. Through 2010 and most of 2011, monetary policies remained fairly loose in Kenya, Uganda and Tanzania, with only cautious and ineffective efforts to tighten, perhaps encouraged by the experience of gradually moderating inflation without policy tightening during the earlier episode in 2009. The commodity price shocks turned out to be much more persistent this time, and they combined with vigorous economic activity, a negative balance of payments shock, and accommodative policy to further accelerate inflation and de-anchor expectations, weakening the exchange rate in an inflationary spiral. Some of the countries began to respond, to varying degrees. In July 2011, Uganda announced a new Inflation Targeting (IT) ‘lite’ policy regime and in August began a still-somewhat gradual tightening of policy. Kenya enacted fitful, partial, and ineffective tightening measures but did clarify its regime in September 2011. In Rwanda, in contrast, tighter monetary policy and a stable nominal exchange rate throughout the period kept inflation from taking off. Finally, the governors of the four central banks agreed at an unusual October 2011 meeting that policy needed to be tightened significantly in order to bring inflation under control, even at a cost to output, and they immediately acted.10 This tightening and surrounding events are our topic here. While the tightening took place in response to economic events, this does not make it entirely endogenous and thus invalidates our narrative approach to identifying the monetary transmission mechanism. Throughout 2011, concerns about the adequacy of the policy stance were increasingly widespread. However, it was unclear when a tightening might come or how strong it would be. Indeed, the narrative suggests that some observers were becoming concerned that it might not come at all. Thus, when it came, it was at least partly unexpected, unusual in the language of Friedman and Schwartz (1963). We can thus ask, what did this large monetary policy tightening shock do? We find some evidence consistent with a clear transmission mechanism. In some of the four countries, after a large policy-induced rise in the short-term interest rate, lending and other interest rates rose, the nominal and real exchange rate tended to appreciate, output tended to fall, and inflation declined. The variation in experiences among the four cases is informative. Most importantly, the cross-country variation in transmission seems to depend sharply on the policy regime in place. In particular, we find the clearest transmission in Uganda, where the IT lite regime itself was simpler and more transparent, and in Kenya, particularly once the authorities explicitly signalled the monetary policy stance with a short-term interest rate and described their intentions in terms of their inflation objective. In regimes where the stance of monetary policy was harder to asses such as Tanzania, which conducted monetary policy under a de jure monetary targeting regime, and Rwanda, a de facto exchange rate peg, the transmission of monetary policy to lending rates and, of course, the exchange rate is less evident. Nonetheless, in Tanzania, the exchange rate seemed to respond strongly to adjustments of the monetary policy stance. We see mixed signs of the importance of financial development. All four countries, like other LICs, have relatively small, concentrated, and bank-dependent financial systems, to varying degrees. In particular, Kenya’s large financial sector makes it an outlier, and it also had perhaps the most complete and unambiguous transmission. However, Uganda, which also clearly demonstrated the main elements of monetary policy transmission, has a relatively small financial sector comparable to that of the other two countries. While the shock was not entirely expected, it was not isolated from outside influences, particularly shocks to global risk appetite and commodity prices, which directly affected exchange rates and prices of traded goods. A close reading of the timing and some statistical evidence suggests that such shocks do not explain most of the exchange rate and price movements around the time of the tightening event, such that the residual unexplained component is consistent with our emphasis on the role of monetary policy itself. Moreover, Uganda’s somewhat earlier policy tightening—starting after its regime change in July—matches an earlier if also somewhat more gradual turnaround in its exchange rate and inflation. The paper proceeds as follows: We first briefly present the stylised facts of the countries under study, including structural features of the economy, the financial system, prices, and the policy regimes. We then proceed to the event study, identifying the policy shock and tracing out the effects of these shocks on the main macroeconomic variables: interest rates, credit aggregates, the exchange rate, output, and inflation. We then analyse more closely the role of important exogenous shocks to capital flows commodity prices that complicate interpretation of the events. Finally, we draw some tentative lessons. 2. The environment for monetary policy 2.1 Structure of the economy and the financial sector The four countries of interest here are in many ways typical of SSA LICs, making their experiences of general interest. Moreover, they share a recent macroeconomic history that is sufficiently stable that the recent monetary policy contraction is a salient event. They are among SSA’s many ‘success stories’ since the mid-1990s, with the achievement of macroeconomic and political stability and rapid economic growth (Table 1). Their economic structure is also broadly characteristic of SSA LICs: low-trade shares, mostly commodity exports, high though falling aid dependence, service-sector-led growth, and a large agricultural sector and rural population. The terms of trade have been fairly stable or rising in recent years.11 Table 1: Basic Economic Indicators, 2011 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Sources: IMF and the World Bank. Table 1: Basic Economic Indicators, 2011 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Sources: IMF and the World Bank. Inflation in the EAC4 is volatile and highly correlated across countries, mostly explained by the high share of food items in the overall CPI. The weight of food prices in the CPI is highest in Tanzania (47%), followed by Kenya (36%), Rwanda (35%) and Uganda (27%).12 Domestic food crops yields in the region are highly dependent on weather patterns, which are characterised by a bimodal annual rainfall cycle. The existence of trade barriers to protect the domestic agricultural production has increased the sensitivity of the domestic food supply to unfavourable weather conditions. Financial systems also share the characteristics that have been associated with weak transmission, though with important cross-country variation. Financial development has been rapid in recent years, but these countries still generally have small and concentrated private-bank-dominated financial systems, a large informal financial sector, shallow capital markets (except Kenya), and short yield curves (except Kenya and more recently Uganda) (Table 2). In all four countries, commercial banks maintain substantial excess reserve deposits at the central bank. Additionally, the four economies exhibit different degrees of financial openness, with Uganda and Kenya being the more financially integrated of the four and Rwanda and Tanzania being the least. Looking across the four countries, Kenya stands out for its relatively developed financial sector, with a larger and less concentrated banking system. Table 2: Financial Sector Indicators, 2011 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Sources: IMF and World Bank. The first four indicators measure aspects of financial depth, while the last column measures financial openness. 1/ Assets of five largest banks as a share of total commercial banking assets. 2/ Index values are for 2010. The index measures the intensity controls over current or capital account transactions, the existence of multiple exchange rates, and the requirements of surrounding export proceeds. It takes a maximum value of 2.5 for the most financially open economies and a minimum of -1.9 for the least financially open. See Chinn and Ito (2008). Table 2: Financial Sector Indicators, 2011 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Sources: IMF and World Bank. The first four indicators measure aspects of financial depth, while the last column measures financial openness. 1/ Assets of five largest banks as a share of total commercial banking assets. 2/ Index values are for 2010. The index measures the intensity controls over current or capital account transactions, the existence of multiple exchange rates, and the requirements of surrounding export proceeds. It takes a maximum value of 2.5 for the most financially open economies and a minimum of -1.9 for the least financially open. See Chinn and Ito (2008). It is not straightforward to characterise the exchange rate and monetary policy regimes of these countries. All four countries reported having floating exchange rate regimes during this period. However, this paper follows the de facto classification in International Monetary Fund (2012), according to which Kenya, Uganda, and Tanzania had floating regimes (with Kenya having only limited foreign exchange intervention and hence labelled a ‘free float’ while the other two at times engaged in substantial intervention and were labelled ‘floats.’) Rwanda was classified as a ‘crawling-peg-like arrangement’ during the period in question (Appendix 1. Table A. 1 reports exchange rate, capital control, and monetary policy regimes for each country during the period in question). Meanwhile, Kenya, Uganda and Rwanda had de jure relatively open capital accounts insofar as there were relatively few official restrictions on capital account transactions, while Tanzania had widespread capital controls in place. As in most SSA countries with de jure floats, all four (except Uganda after July 2011, which labelled its regime ‘IT lite’ after that point) conducted monetary policy under a de jure monetary aggregate targeting framework, in principle adjusting the money supply to achieve intermediate targets in terms of broad money growth.13 However, these three regimes were much more flexible, and complex, in practice (Appendix 1. Table A.1). In particular, the experience of these countries matches the broader experience with such de jure money targeting regimes, which is that target misses are frequent, and at least in relatively low-inflation environments such deviations are not associated with misses of inflation objectives. Rather, central banks tend to make judgments on an ongoing basis as to whether targets should be achieved, and if not how they should be revised for the next quarter, depending on outcomes in money and exchange rate markets and a broader sense of whether inflation and output (and other) objectives are being achieved. This is for various reasons, not least because to adhere to the targets would generate excessive volatility of short-term interest rates in the face of money demand shocks. This makes the stance of policy hard to grasp and in particular very hard to infer from monetary aggregates themselves.14 Tanzania and especially Rwanda adhered most closely to their de jure money targeting regime, though even here economically meaningful deviations were frequent.15 Kenya had over time paid less and less attention to monetary aggregates, culminating in September 2011 with a clear announcement of a move to use the short-term interest rate as its main policy instrument, with the objective of achieving its inflation objectives (see Andrle et al. 2013). Uganda too had undergone an important evolution in this regard, moving in October 2009 from quite strict to flexible money targeting with substantial attention to interest rates as operating target to an explicit ‘inflation targeting ‘lite’ regime in July 2011 with use of short-term interest rates in pursuit of its inflation objectives. In sum, Kenya (especially after September 2011) and Uganda (especially after July 2011) adhered reasonably closely to an IT-style regime, with a managed float, a reasonably open capital account, interest rates as monetary policy instrument, and inflation stability as the clearly stated objective of monetary policy. Tanzania had a more complex regime, with a more heavily managed float, a multiplicity of monetary policy instruments and intermediate targets, and a less clear statement that inflation stability was the main objective of monetary policy. Finally, Rwanda seems to have largely subordinated its monetary policy to its de facto crawling peg exchange rate, though the de facto somewhat closed capital account may have given it some room for manoeuvre on monetary policy as well. The different degree of attention to monetary aggregates in practice has corresponded to varying degrees of interbank interest rate volatility. This volatility is clearest in Tanzania and least in Uganda and Kenya, consistent with their de facto use of interest rates as operating targets and indicators of the stance of policy. In Rwanda, lack of market clearing even in the interbank market seems to have allowed some degree of disconnect between money aggregates and short-term interest rates. In these hybrid regimes, it is often difficult to know how to interpret any particular interest rate. Sometimes ‘policy rates’ are not necessarily market clearing, present no arbitrage opportunities with other short-term interest rates, and may contain no signal of policy intention. But this can change suddenly as the details of central bank operations change. Pressures from fiscal authorities can encourage these central banks to create deviations between ‘policy rates’ and some market rates, particularly those at which the Treasury finances its activities, sometimes creating further opacity with respect to interest rates. In assessing the stance of policy, we in general refer to the interbank rate as an indicator of the market short-term interest rate, with due reference to ‘policy rates,’ exchange rate interventions, and money aggregates as appropriate. Other instruments, such as changes in reserve requirements, may have some independent signalling import but also are likely to work at least in part through their influence on interbank rates. 3. The event study We define a shock as an episode in which a central bank undertakes overt and unusual—large and substantially unexpected—actions to exert a contractionary influence on the economy in order to reduce inflation. The tightening we consider here took place mainly in October 2011 around the time of a meeting of EAC Central Bank governors at which it was stated that inflation was getting out of control and that monetary policy needed to be tightened. This meeting and the resulting sharp policy actions represent the distinct monetary policy shock that allows us to trace the transmission mechanism. We now take a closer look at the tightening episode. 3.1 The run-up Through mid-2011, international prices of food and energy shot up by more than 30% and 40%, respectively (Figure 1, Panel 1). Meanwhile, economic activity in the EAC4 was generally recovering from the earlier effects of the global financial crisis (Panel 2). Direct evidence suggests that the monetary policy stance was mostly accommodative, with nominal rates fairly flat and real interest rates mostly negative in all three countries (Panels 3 and 4). Partly reflecting this policy stance, but also at times owing to pressures on the capital account from swings in global risk aversion, nominal real exchange rates were generally weakening (Panels 5 and 6).16,17 Figure 1: View largeDownload slide The Run-Up. Sources: IMF estimates, Haver, national authorities. See footnote 19 for detailed calculation. Figure 1: View largeDownload slide The Run-Up. Sources: IMF estimates, Haver, national authorities. See footnote 19 for detailed calculation. By 2011Q3, these factors were reflected in headline inflation in Kenya, Uganda and Tanzania that surpassed the common inflation target of 5% (Panel 7).18 Even though most of the increase in headline inflation during this period is explained by the acceleration in food and fuel inflation, core inflation also increased substantially in all countries, almost doubling during the course of a year, and in all cases considerably overshooting the common inflation target (Panel 8).19 Consistent with the negative real interest rates, the monetary policy authorities were generally ‘behind the curve’ in responding to the building inflationary pressures.20 Moreover, policy responses were generally poorly signalled both in terms of the statements of the authorities and in that different instruments, such as different short-term interest rates, gave different signals. As inflation in Kenya continued to increase and the nominal exchange rate depreciated by about 10% in the period from March until September 2011, the Central Bank of Kenya (CBK) responded fitfully and opaquely. For example, a March increase of the Central Bank Rate (CBR) by 25 basis points to 6%, reversing a lowering of 25 basis points in January 2011, was accompanied by mixed signals as to the intent of the CBK, and had no discernible effect on lending rates or the exchange rate.21 In August 2011, the central bank resorted to stronger moves, including restricting access to the discount facility and restricting liquidity provision through open market operations, still keeping the CBR unchanged. These moves resulted in a brief intra-month 2,200 basis point spike in interbank rates, which had little apparent effect on treasury bill rates and none on lending rates or the exchange rate, which continued to depreciate. In September, the CBK clarified its operating regime, emphasising that it would use the CBR as its main policy instrument with the objective of achieving its inflation objectives. However, while the central bank raised the CBR from 6.25% to 7%, at the same time, and in contradiction, it provided liquidity to the interbank market at 5.75%. Meanwhile, policy statements from the CBK were ambiguous and lacked a clear announcement of policy tightening.22 Tanzania began acting as early as 2010Q4, by some measures. A sharp contraction in the real growth rate of money in late 2010 caused a jump in the interbank rate, with a hint of pass-through to T-bill rates, but this proved short lived even as the growth rate of money continued to slow. As real money growth reached nearly zero in mid-2011, interbank (but not T-bill) rates again spiked. There was no discernible effect of these actions on lending rates, the exchange rate, or credit. Uganda, in contrast, began effective tightening earlier. Interbank rates increased by some 500 basis points during the first half of 2011, though with little apparent effect; for example, lending rates remained unchanged. The BOU tightened much more consistently and coherently after the July 2011 announcement of a new ‘IT-lite’ regime and the introduction of the CBR to signal the stance of policy.23 The central bank raised the CBR in two steps from 13% in July to 16% in September. In this case, the lending rate increased by 340 basis points from July until September 2011, and the exchange rate began to stabilise, with the real exchange rate appreciating in September. We return to the effects of these measures in the next section. Rwanda presents a contrasting picture in terms of its policy regime and stance: through 2010–2011 output was below trend, the real exchange rate and real interest rates were stable (and the latter were positive). This may have reflected a tighter monetary policy stance, as well as the de facto crawling peg exchange rate regime. It is hard to infer much about the stance of policy during this period from the monetary aggregates; for example, real reserve money growth varied sharply in a way that seems unrelated to short-term interest rates and the exchange rate or for that matter inflation and the output gap. 4. The event and its aftermath As 2011 unfolded, inflation accelerated on the strength of higher food and oil inflation, strong demand, weakening exchange rates, and still-negative interest rates (Rwanda is the exception). The EAC4 came to the realisation by their October meeting that action was needed to stabilise the situation. On October 5th, the CBK increased the CBR by 400 basis points and on November 1 by a further 550 basis points to 16.5%. During this period, it also increased the cash reserve requirement by 75 basis points to 5.25%, and adjusted the discount window rate by more than 20%, ‘as decisive and immediate action is required from the monetary policy side to stem these inflationary expectations.’24 In addition, the Ministry of Finance lowered the foreign exchange exposure limit for commercial banks to 10% of core capital from 20%.25 The Bank of Uganda (BUG) followed up early substantial tightening, and the introduction of its new regime in July, by raising its policy interest rate by 400 basis points in October and a further 300 basis points in November to 23% and stepped up its intervention in the foreign exchange market to contain the depreciating pressures on the Shilling, stating that ‘the upside risks to inflation have increased, it is necessary to tighten monetary policy further… (this) should be seen as a clear signal of the BOU’s determination to bring inflation under control. However, should the upside risk to inflation continue in the months ahead, then monetary policy will be tightened further.’26 In Tanzania, the central bank increased its policy rate by 200 basis points in October and a further 202 basis points in November to 11.00%, and on October 26 augmented the minimum reserve requirements on government deposits held by banks from 20% to 30%, reduced commercial banks’ limit on foreign currency net open positions from 20% to 10% of core capital, tightened capital controls, and increased sales of foreign exchange in the interbank market.27 This shift was more decisive than earlier efforts, perhaps partly because this time the authorities put more emphasis on the policy rate, as well as on the quantitative actions. Finally, the National Bank of Rwanda took a variety of much more moderate tightening measures, consistent with the much lower degree of disequilibrium throughout 2011. Again, it is hard to point to any specific measurable action with respect to money aggregates, but the central bank’s policy rate was increased by 50 basis points to 6.5%, as ‘the Central Bank finds it appropriate to review its policy rate in order to keep the monetary aggregates at optimal levels to limit inflation pressures while continuing to support economic growth.’28 Having identified the policy-tightening event and the variations across the four countries, we now turn to an assessment of the various channels of transmission of monetary policy across our group of countries. The variety of instruments and (time-varying) differences in regimes can make it difficult to characterise in a simple way the stance of policy itself. We generally use the interbank rate as the best single measure of the policy shock itself. As a measure of the short-term market rate in all four countries, it captures aggregates the effects of policy rates and quantitative policies (quantitative interventions, reserve requirements) and distills the divergent effects of potentially inconsistent actions taken with various not-necessarily-market-clearing ‘policy rates.’ 4.1 The interest rate channel In Uganda and Kenya, the pass-through from policy rates to interbank rates was fast and complete (Figure 2, Panels 1 and 2) (and to T-bill rates, shown in the working paper). Tanzania’s battery of measures also transmitted quickly to money market rates. Figure 2: View largeDownload slide The Monetary Policy Contraction and its Aftermath. Source: IMF estimates, Haver, national authorities. Figure 2: View largeDownload slide The Monetary Policy Contraction and its Aftermath. Source: IMF estimates, Haver, national authorities. The relevance of the policy regime is evident in the transmission to banking rates (Panel 3). Lending rates, in particular, responded swiftly—though partially—to the monetary policy contraction in Kenya and Uganda. Uganda’s lending rates began to respond somewhat earlier, corresponding to the August/September tightening. There is little sign of transmission to lending rates in Tanzania and Rwanda. The lack of response of lending rates in Tanzania and Rwanda, despite the increases in the interbank rates in these two countries, is reminiscent of the non-response of lending rates to Kenya’s August spike in interbank rates. Even in Kenya and Uganda, the pass-through from short-term market to lending rates was partial. This may reflect lack of competition or other structural weaknesses in the financial system, as argued in Montiel et al. (2012). However, lending rates are longer-term rates, so partial pass-through to a (at least somewhat temporary) tightening is also consistent with fully functioning markets, by which we mean arbitrage across returns of different assets, and an expectation that the period of high short rates will be somewhat shorter than the tenor of the loans.29 Moreover, standard data in these countries (such as we use here) report average, not marginal lending rates. Finally, as we have already argued above, the pass-through of policy rates to short-term market and lending rates will depend on the clarity of the regime and in particular the ability of market participants to infer that the increase reflects policy intent and will not be quickly reversed.30 It may be that there was still some uncertainty about whether the authorities in Kenya and Uganda would stay the course. 4.2 The bank lending channel The data indicate the existence of the credit channel in Kenya, Uganda, and Tanzania. In these three countries growth in credit to the private sector peaked soon after the policy contraction started and decelerated substantially as the monetary authorities stepped up the pace of tightening. Accordingly, during the 2011Q3 to 2012Q3 period, credit to the private sector growth in Kenya, Uganda and Tanzania decelerated. Again, Uganda’s contraction began a month or two before the others. There are also signs of credit rationing in the case of Tanzania: even though lending rates did not respond to the tightening, there was a meaningful impact on the quantity of credit extended to the economy. In Rwanda, there is little sign of slower credit growth, perhaps reflecting the much less significant tightening. Abuka et al. (2015) provide important supportive evidence for the bank lending channel in the case of Uganda. A unique data set of loan-level data spanning the period in question allows them to control for demand effects through region-industry dummies and aggregate variables such as GDP, directly. This allows them to interpret the effects of the policy shock as causal for credit supply and as not due to demand effects. They find that higher short-term market rates interest rates are associated with an increase in banks’ lending rates and reductions in loan volume at the extensive and intensive margins. The strength of this bank lending channel is significant, albeit about half of that observed in advanced economies studied with similar data and techniques.31 4.3 The exchange rate channel In Kenya, Uganda, and Tanzania, the increase in short-term interest rates was associated with a contemporaneous appreciation of the currency. Notably, this took place during the decisive tightening phase in 2011Q4, but not earlier when policy was more cautious and less clearly signalled. Uganda is again a partial exception insofar the exchange rate stabilisation began two months earlier, corresponding to its earlier tightening phase. Rwanda’s de facto pegged regime, relatively closed capital account, and more moderate tightening are apparent in the absence of large exchange rate movements.32 4.4 Output The tightening episode is associated with a contraction of output in Uganda and to a lesser extent in Tanzania (Panel 7). The absence of a visible decline in the output gap in Kenya is notable. Of course, factors other than monetary policy such as fiscal policy and foreign demand also influence the output gap, and it is difficult to measure the output gap in the context of frequent supply shocks.33 The Abuka et al. (2015) analysis based on loan-level data provides supportive analysis of the effects of this particular monetary policy shock on output in Uganda. By identifying differential loan supply effects in districts with varying banking sector conditions, and measuring real effects through night-time light output measured from satellites, they can plausibly identify the real effects of monetary policy acting through the bank balance sheet channel. They find that output does indeed contract more in those districts where banks have balance sheets, suggesting strong balance sheet channel for the monetary contraction. 4.5 Inflation Finally, the inflation rate came down sharply with the monetary contraction. Headline inflation began turning around rather quickly, within a month or two, presumably reflecting the rapid pass-through of exchange rate movements. The turnaround was sharpest in Uganda and Kenya but was apparent in Tanzania as well. Again, Rwanda shows a much more gradual pattern, reflecting the fact that inflation was never far from target. Core inflation followed, albeit much more gradually (Panel 9). 5. High-frequency and statistical evidence on the independent role of monetary policy As the above narrative suggests, the monetary policy shock of late 2011 was not entirely expected, but it was also not isolated from outside influences, particularly shocks to global risk appetite and commodity prices. Because these shocks directly affected exchange rates and prices of traded goods, as well as monetary policy, they may have led us to overemphasise the importance of monetary policy shocks per se. In particular, an alternative story emphasises that inflationary pressures were not a reflection of vigorous economic activity and loose monetary policy but a result of the higher food and fuel prices as well as global risk premium shocks that depreciated exchange rates. And along the same lines, the eventual reductions in inflation reflect the fading of the effects of these global shocks, rather than the monetary policy response. In this section, we take two approaches to desentangling these factors. First, we take advantage of the availability of high-frequency (daily) data on global risk shocks and on key associated endogenous variables (the exchange rates and domestic interest rates) to look closely at the timing of these shocks. In particular, we examine whether monetary policy shocks that are not coincident with risk premium shocks are correlated with movements in the exchange rate, and whether the sign of this correlation is consistent with the narrative above or rather with reverse correlation from the exchange rate movements to interest rates. To summarise, we find that increases in short-term domestic interest rates around the times associated with the monetary policy tightening are correlated with appreciation of the exchange rate and generally not with coincident improvements in global risk appetite, providing direct evidence in favour of the above narrative. Second, we. large movements in the exchange rate are associated with risk premium shocks or with can be associated with monetary policy shocks per se (as measured by the interest rate) or w. We also look at This evidence suggests that such shocks do not explain most of the exchange rate and price movements around the time of the tightening event, such that the residual unexplained component is consistent with our emphasis on the role of monetary policy itself. Moreover, Uganda’s somewhat earlier policy tightening—starting after its regime change in July—matches an earlier if also somewhat more gradual turnaround in its exchange rate and inflation. 5.1 A high-frequency analysis of the role of shocks to role of global risk appetite Are shifts in capital flows and global risk aversion able to provide an alternative explanation of the exchange rate dynamics during the run-up and following the coordinated tightening? The year 2011 was one of increased global risk aversion, with the rising political tensions in the Middle East associated with the Arab Spring, the sovereign debt crisis in Europe, and the downgrading of the credit rating of major industrial economies. This surely contributed to exchange rate pressures on the EAC4, but cannot plausibly explain the timing and magnitude of the real depreciations observed. The currencies of Kenya, Uganda, and Tanzania started to weaken in 2010, well in advance of the episode, standing during 2010 and 2011 amongst the most depreciated currencies in the emerging and frontier markets world (excluding fixed exchange rate regimes, Figure 3). Figure 3: View largeDownload slide Emerging and Frontier Markets Real Exchange Rates. (Real effective exchange rate based on CPIs, Jan. 2010 = 100, increase means appreciation35). Sources: IMF INS database and authors’ computation. Figure 3: View largeDownload slide Emerging and Frontier Markets Real Exchange Rates. (Real effective exchange rate based on CPIs, Jan. 2010 = 100, increase means appreciation35). Sources: IMF INS database and authors’ computation. Taking a closer look at the high-frequency data, swings in global risk appetite can partly explain the sharp depreciation during the July–September period and perhaps some of the appreciation that followed the monetary policy tightening. However, the timing of the turnaround indicates a strong independent role for the monetary policy contraction. Global risk appetite, as proxied by the VIX Index, deteriorated markedly in August and September as the credit ratings of the United States, Japan, and Italy were downgraded and Europe’s debt crisis intensified (Figure 4). Tensions in international capital markets eased by late September. However, the currencies of Kenya and Tanzania reached their lowest levels weeks later.34 The Kenyan shilling strengthened immediately on the second day of the policy tightening announcement, after staying near low levels despite improved global sentiment. The Tanzanian shilling, on the other hand, continued to weaken even after the announcement of the coordinated policy, interrupting its slide only after a further battery of measures was announced on the week of October 24. Figure 4: View largeDownload slide VIX (Index) and EAC4 Exchange Rates. Source: Bloomberg. Note: Exchange rates are local currency units/dollar; an increase indicates a nominal depreciation. Figure 4: View largeDownload slide VIX (Index) and EAC4 Exchange Rates. Source: Bloomberg. Note: Exchange rates are local currency units/dollar; an increase indicates a nominal depreciation. This evidence suggests a role for monetary policy per se in driving the exchange rate, in that the exchange rate appreciations coincided with the monetary policy tightening much more closely than with the improvements in international risk sentiment, when we look closely. 5.2 The role of global shocks: a statistical counterfactual In this section, we construct a statistical counterfactual to examine directly the view that global shocks to financial conditions and commodity prices are sufficient to explain the rise and subsequent fall in inflation and other main macroeconomic dynamics around the events in question. While we do not think there are enough data and stable enough regimes to identify statistically the effects of monetary policy through this episode, we can estimate the effects of these global (and thus exogenous) shocks on inflation and other key variables in the countries of interest here. By omitting monetary policy from the regressions, we implicitly attribute any effect of the coincident monetary policy response to these global shocks to the global shocks themselves. Thus, the fitted values from these regressions represent a counterfactual that assumes no explicit or unexpected monetary policy response to the global shocks, or that such a response has no independent effect on macroeconomic variables. Before we carry out this exercise, two points are worth noting. First, as shown earlier, core inflation (mainly excluding food and energy prices) also increased substantially through 2011, almost doubling and overshooting the target, before beginning to come back down, Second, movements in food and fuel inflation are themselves influenced by the monetary policy stance. Monetary policy influences the domestic price of imported food and fuel through the exchange rate. For locally produced (and not-fully-traded) food, monetary policy can work through aggregate demand. Thus, disentangling the contribution of monetary policy from that of supply shocks cannot solely be based on separating food/fuel and core inflation. Andrle et al. (2013) discuss these issues and apply a simple structural model to events in Kenya during 2011 and find that monetary policy accounts for much of the inflation dynamics, including the behaviour of domestic food prices. We estimate two simple regression models, explaining actual nominal exchange rate depreciation and inflation with only clearly exogenous variables. The identification strategy is to give the maximum possible influence to these variables, by interpreting any comovement with domestic policy variables such as the interest rate as being due to these exogenous variables. For exchange rate depreciation, the exogenous are the percent changes in the VIX and the Fed funds rate. For CPI inflation, the exogenous variables are US inflation and world food and oil prices. We assume an AR(2) process as follows: Δyt=c0+c1Δyt−1+c2Δyt−2+cXt+εt Where Δy is the endogenous variable (nominal exchange rate depreciation or inflation) and Xt is a matrix of exogenous explanatory variables. We use monthly data on nominal exchange rate depreciation and inflation in Kenya, Rwanda, Tanzania and Uganda. Both nominal exchange rate depreciation and inflation are month-on-month annualised changes, with the nominal exchange measured in local currency units per US Dollar. Spectral regression is used to estimate model parameters. This is equivalent to pre-filtering the data to include only selected frequencies and then applying OLS to this filtered data. This enables us to infer model parameters from the business cycle dynamics in the data while mitigating adverse effects of very high-frequency noise. This regression approach provides more robust parameter estimates, e.g., parameters remain relatively robust to changes in the data range. Furthermore, compared to standard OLS the spectral regression boosts effects of exogenous variables and increases the explanatory power of the regression, as measured by R2 statistics. Estimation outcomes are presented in Tables 3 and 4 and are inferred from cycles with periodicity of 4–96 months.36 Table 3: Regression Results—Nominal Exchange Rate Depreciation Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Source: Authors’ computation. The ‘US Rate’ is the Federal Funds rate. Note: Statistically significant parameters are denoted by stars (***, **, and * denote 99%, 95% and 90% confidence level). The ‘US Rate is the Fed Funds rate. The exchange rate is the nominal exchange rate measured in local currency units/US Dollar. Table 3: Regression Results—Nominal Exchange Rate Depreciation Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Source: Authors’ computation. The ‘US Rate’ is the Federal Funds rate. Note: Statistically significant parameters are denoted by stars (***, **, and * denote 99%, 95% and 90% confidence level). The ‘US Rate is the Fed Funds rate. The exchange rate is the nominal exchange rate measured in local currency units/US Dollar. Table 4: Regression Results—Inflation Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Source: Authors’ computation. Note: Statistically significant parameters are denoted by stars (***, ** and * denote 99%, 95% and 90% confidence level). Table 4: Regression Results—Inflation Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Source: Authors’ computation. Note: Statistically significant parameters are denoted by stars (***, ** and * denote 99%, 95% and 90% confidence level). We then construct counterfactual measures using only the exogenous variables along with the associate endogenous dynamics. The counterfactual measures represent what would be dynamics of endogenous variables without monetary policy shocks. The counterfactual measures are generated starting in 2011M04. The regression explains a fair amount of the variance of the exchange rate (from 56% in Uganda to 46% in Rwanda, presumably low owing to its quasi-managed regime). Most of this explanatory power is due to the importance of the lagged exchange rate itself, though the exogenous variables are highly significant. In Figure 5, we plot the predicted value of the exchange rate based only on the exogenous shocks and associated endogenous dynamics (the ‘counterfactual,’ which we interpret as reflecting the dynamics of the exchange rate in the absence of unexpected monetary policy decisions).37 As Figure 5 shows, in all but Rwanda the exogenous factors go in the right direction in explaining the depreciations in Q3 and subsequent appreciations, but much less than observed and with not quite the right timing. The peak depreciation of the predicted exchange rate in all countries occurs around September, but the actually (much weaker) bottom occurs in October in Tanzania and Kenya and in August in Uganda, more consistent with the timing of the monetary policy shock.38 Figure 5: View largeDownload slide Nominal Exchange Rate Depreciation, MoM Annualised in Percent. Source: Authors’ computation; see Table 3. Figure 5: View largeDownload slide Nominal Exchange Rate Depreciation, MoM Annualised in Percent. Source: Authors’ computation; see Table 3. The results for headline inflation paint a similar story. Inflation peaks coincide or follow with a one-month lag in the nominal exchange rate depreciation and they cannot be explained by the counterfactual (Figure 6). Figure 6: View largeDownload slide CPI Inflation, MoM Annualised in Percent. Source: Authors’ computation; see Table 4. Figure 6: View largeDownload slide CPI Inflation, MoM Annualised in Percent. Source: Authors’ computation; see Table 4. To summarise this section, the swings in global risk aversion and the food and fuel supply shock during 2011 did play an important role during the episode under study, they are only part of the story, both in terms of magnitude and timing, and do not overturn the conclusion that monetary policy seems to have played a decisive role. 6. Summary and interpretation We have identified a moment when three of the EAC4 broke from previous behaviour and executed more-or-less clearly signalled monetary policy contractions with the explicit intent of reducing inflation. We find clear evidence of most elements of the standard transmission mechanism in most of the countries. The transmission was the clearest in Kenya and Uganda, where market and lending rates followed the policy rate with little lag, the nominal exchange rate appreciated sharply on the policy announcement, credit growth (and, at least in Uganda, the output gap) began to decline immediately. Both headline and core inflation also began to decline almost immediately. Transmission was less clear in Tanzania, where the effects on some interest rates, activity, the exchange rate, and inflation are still broadly evident, but lending rates failed to respond and the effects on output are barely evident. Rwanda presents a control along several dimensions: initial imbalances were much smaller, the tightening much less significant, and the various components of transmission much more muted or invisible. Based on and summarising the preceding narrative, we can now evaluate the two hypotheses discussed in the instruction to explain the variation in cross-country experience. In the episode under study, we find substantial explanatory power in the idea that the nature of the policy regime conditions transmission. In the cases of clearest transmission, Kenya and Uganda, the regimes in October 2011 most resembled inflation targeting in that the authorities prioritised inflation, emphasised the role of the policy rate, allowed the nominal exchange rate a large degree of flexibility, and broadly avoided multiple objectives. Earlier tightening efforts by Kenya, e.g., in August 2011, were more incoherent in terms of the consistency across different instruments and communications and did not translate into lending rates or the exchange rates. Uganda’s earlier tightening efforts were more coherent and stronger following its July 2011 move to ‘IT lite’ in July and indeed had some effect on lending rates, credit, the exchange rate and inflation about two months before Kenya and Tanzania. In Tanzania instead, the money targeting regime led to highly volatile short-term interest rates, a variety of instruments were used in not-always-consistent ways, and overall there was less clear signalling of the policy stance. Rwanda’s regime was in some ways the most complex, with a quasi-pegged exchange rate, direct influence on private sector credit, direct influence on private sector credit, monetary aggregate targets and a policy rate. The emphasis on the exchange rate left room for monetary policy itself to act, and insofar as it could, the regime did not provide a clear signal. In the event, there was apparently less tightening, and less need to tighten. The second hypothesis is that transmission worked better in countries with greater financial depth and more open capital accounts. We find mixed support for this story in this episode. It remains plausibly the case that countries with more liquid and deeper financial markets will observe stronger transmission from policy rates to the macroeconomy.39 However, in this particular case measures of financial depth do not seem determinative for the clarity of transmission. A glance back at Table 2 reminds us that Kenya is the clear outlier for all measures of financial depth, with the other three countries remarkably similar. And yet, as we have argued, the evidence for transmission looks fairly strong, and similar, for Uganda and Kenya, in contrast to Tanzania and Rwanda. On the other hand, the narrative is consistent with the view that the lower degree of capital account openness in Tanzania and Rwanda (Table 2) may have contributed to obscuring or impairing transmission in these two countries. The exchange rate did seem to respond in Tanzania, but less dramatically than in Kenya and Uganda. Finally, it is often asserted that the high levels of excess reserves usually observed in SSA countries prevent the operation of the monetary transmission mechanism, for example because tightening policy may amount to ‘pushing on a string’ as banks respond to a contraction by withdrawing excess reserves.40 In the episode we examine here, excess reserves did not seem to impair the transmission mechanism. As Figure 7 shows, there are indeed substantial reserves in excess of required levels in all four countries, with large variations across time and countries but with no evident influence on transmission. While in Kenya, excess reserves fell during 2011, they remained above 6% of required reserves even in October, and they rose in Uganda and varied around 20% of required reserves during the peak of the tightening phase, higher than they had been since 2007.41 Figure 7: View largeDownload slide Excess Reserves. Source: Central Bank Data. Figure 7: View largeDownload slide Excess Reserves. Source: Central Bank Data. 7. Conclusions The identification of monetary policy transmission is difficult under any circumstances, and especially so in countries with poor data, obscure and time-varying policy regimes, and frequent supply shocks. As emphasised by Summers (1991) among others, the analysis of dramatic events such as the Great Depression and the Volker disinflation in the United States has played a critical role in forming professional opinion and framing the discussion in advanced countries. However, such analyses are scarce in developing countries. We have taken advantage of a dramatic tightening of monetary policy in four countries in East Africa in October 2011 to trace through the effects of this tightening on interest rates, credit, the exchange rate, output, and inflation. We find clear evidence of a working transmission mechanism in two of the countries: after a large policy-induced rise in the short-term interest rate in Kenya and Uganda, lending rates rose, the exchange rate appreciated, credit growth slowed, output growth tended to fall, and inflation declined. The other two countries represent a contrast to varying degrees. In Tanzania, some but not all market rates and the exchange rate seems to respond to adjustments of the monetary policy stance, credit growth slowed, and we see some signs of the effects of policy in output (possibly through credit rationing); in Rwanda, the initial disequilibrium was much smaller, the exchange rate more-or-less controlled, any monetary policy tightening much less evident, and any effects of such a policy tightening as took place therefore difficult to observe. These case studies provide many illustrations of the role that the policy framework itself plays in governing the strength of transmission. Most importantly, Kenya and Uganda by October had clarified their inflation objective and the centrality of the policy rate as the main signal. In this context, they were able to clearly articulate that they were raising rates to bring inflation back down. Earlier efforts in Kenya in August, before this clarification, were ineffective. Tanzania represents an intermediate case, in which a continued focus on money targeting and some inconsistency across policy instruments coexisted with a somewhat less clear transmission. The proliferation of policy instruments was common across all four countries. This put a premium on coherence and communication in signalling the policy stance. The difficulty in interpreting the different measures of policy may impact not only our ability as researchers to discern events but the capacity of interest rates to clear markets and signal policy. From a methodological point of view, this active use of a wide set of instruments under differing policy regimes with multiple objectives, along with the suggested quasi-contemporaneous nature of the transmission mechanism suggest that extra care should applied used when using standard statistical procedures, such as VARs, to measure the effects of policy, especially when those studies are conducted for country groups. In contrast, we found some role for financial openness but little for financial depth or the degree of excess reserves in explaining the cross-country patterns of transmission in these particular cases. We should not overemphasise or over-generalise this result. It remains plausible that countries with less developed financial markets and less open capital accounts will observe weaker transmission. However, this does not seem to have been a determinative feature here. Clearly, in general, shocks other than those to monetary policy are the main drivers of macroeconomic outcomes in countries such as those we examine here. And here as elsewhere, monetary policy was not made in a vacuum, and identification of shocks and their effects is challenging. It is always a great leap of faith to suppose that a residual in an estimated monetary policy reaction represents such a shock and not a misspecification (e.g., an omitted variable or a nonlinearity) in the reaction function. In our cases, we have used the historical narrative to attempt to argue that much of the shock was a surprise, and we have tried to exclude simple alternative hypotheses about the drivers of some of the key variables. But ultimately the results cannot be definitive. Much remains unknown about the transmission mechanism in these countries. The role of supply shocks, food prices, the banking system and limited financial participation, fiscal policy, limited capital account openness, and many other features deserve further exploration. The transmission mechanism appears to be working, particularly when signals are clear and the regime simple and coherent. Acknowledgements This paper is part of a research project on macroeconomic policy in low-income countries supported by the U.K.’s Department for International Development. The project draws in part on earlier work with IMF colleagues, particularly Rogelio Morales. We would also like to thank Chris Adam, Steve O’Connell, and seminar participants at the IMF’s African Department and the Center for the Study of African Economies in Oxford. All errors remain ours. Footnotes 1 See Mishra et al. (2012). 2 Li et al. (2016). For reviews of the literature, see International Monetary Fund (2008), Mishra and Montiel (2013), and Davoodi, Dixit and Pinter (2013) for the East African Community. 3 Li et al. (2016). 4 Romer and Romer (1989), pp. 1. 5 The exogeneity of Romer and Romer’s policy events is challenged in Shapiro (1994) and Dotsey and Reid (1992). See the discussion in Christiano et al. (1999). 6 We thank an anonymous referee for suggesting this interpretation of these data. 7 By ‘regime’ we mean the monetary policy and exchange rate regime, which are of course related. Textbook classifications involve pegs and floats, and money or inflation targeting. As described in more detail below and in Appendix I, the reality in these countries is substantially more complex. We include a discussion of capital mobility in characterising the policy regime, since it also conditions the role of monetary policy. 8 See O’Connell et al. (2015). Cottarelli and Kourelis (1994) show that low pass-through to lending rates can be attributed to the volatility of money market rates, which limits their information content regarding the stance of monetary policy and which is characteristic of the quantity-based monetary frameworks in place in these two countries. The presence of a high level of noise in money market rates can also hinder the degree of money market development, another structural feature identified in the literature as relevant in explaining the transmission of monetary policy impulses to lending rates. 9 Of course, these two sets of hypotheses need not be mutually exclusive. Montiel et al. (2012) emphasises both structural (financial underdevelopment) and regime-related (de facto pegged exchange rates) reasons for attenuated transmission. 10 In the communique from the October 12, 2011 meeting (available at ‘https://www.bou.or.ug/bou/bou-downloads/press_releases/2011/Oct/EAC_GOVERNORS_MEETING_ON_THE_CURRENT_CRISIS_-_OCTOBER_12_2011.pdf’), the ‘Governors observed that the region is facing very high inflation originating primarily from high food and fuel prices but also from demand pressures. The Governors also observed that the region is facing pressures on the currencies to weaken and exchange rate volatility. The pressures for the currencies to weaken result mainly from the widening of the current account deficit originating from rapid expansion of the oil import bill and imports for infrastructure development. In addition, the exchange rate volatility has been due to the effects of the Euro sovereign debt crisis and currency speculation activities. Given these challenges, the Governors agreed to coordinate the following actions: Tightening monetary policy, Stemming volatility in the foreign exchange markets, and Curbing currency speculation activities.’ The motivations for acting jointly presumably included the benefits of political solidarity in the context of the drive for economic union in the East Africa Community, a fear that disjoint action might result in undesired exchange rate fluctuations within the group, and a view that common action might have a stronger signalling effect. 11 Berg et al. (2013) presents much more background information and references. International Monetary Fund (2015) surveys the monetary policy environment, regimes and policy challenges in a broader context. 12 The Tanzania CPI survey includes rural households, unlike the surveys in Kenya, Uganda, and Rwanda. 13 ‘IT lite’ is a term coined (as far as we are aware) in Stone (2003), who characterise it as a regime in which countries that float their exchange rate and announce an inflation objective but are not able to maintain the inflation target as the foremost policy objective. It has come to be used to describe countries that adhere to something like ‘inflation targeting (IT)’ but without all the elements often considered to characterise such a regime. 14 See International Monetary Fund (2015),Appendix 2, Adam et al. (2010), and Kasekende and Brownbrige (2011) for discussions of this issue and Andrle et al. (2013) for Kenya specifically. 15 As shown in International Monetary Fund (2015), deviations that occur roughly once per year correspond in magnitude, by back-of-the-envelope calculations, to interest rate deviation of 5 percentage points in Rwanda and 20 percentage points in Tanzania. Rwanda’s close control of the nominal exchange rate created some tensions with the money targets in practice. 16 Section IV.G examines the role of exogenous external shocks more closely to disentangle them from that of policy. 17 Rwanda was an exception to these generalisations, with output below potential (though rising), a stable nominal exchange rate, and positive real interest rates. 18 Output gaps are estimated with a Hodrick-Prescott filter on the 4-quarter cumulative real GDP in Uganda and Tanzania and Non-Agricultural GDP in Kenya and Rwanda. The estimation sample includes data up to 2012Q3 to correct for end-of-sample bias. Real interest rates are calculated using the 12-month backward-moving average CPI-based inflation rate. The nominal exchange rates are bilateral with the US dollar. The real exchange rates are CPI-based bilateral rates with the US dollar. 19 To assess non-commodity inflationary pressures in a comparable way across the EAC4, we constructed an indicator of core inflation excluding food and fuel prices. These measures are not necessarily the same as those compiled by local authorities. For details, see Berg et al. (2013). 20 In part, this may have reflected the experience of the earlier episode of external price shocks in 2007/2008, when temporarily rising inflation was followed by the commodity price and external demand collapse of the global financial crisis, obviating the need for a monetary policy response in that case, as discussed in Berg et al. (2013). 21 In March 2011, the CBK suggested somewhat contradictorily that ‘this tightening will provide a solution to inflationary pressure and will stabilize the exchange rate while still protecting economic activity’ (Central Bank of Kenya (2011a)). In July, they still considered this action sufficient to mitigate soaring inflation (Central Bank of Kenya (2011b)). 22 At an extraordinary meeting on September 14, 2011, the CBK announced that ‘The high overall inflation environment is mainly a consequence of high food prices and high fuel and energy prices….The CBK will pursue the inflation objective through a continuation of the gradual tightening of monetary conditions.’ (Central Bank of Kenya (2011c)). 23 With the introduction of the IT-lite framework also came the release of a Monetary Policy Statement, signed by the governor and providing forward guidance to market participants. 24 Central Bank of Kenya (2011d). 25 In October 2011, the CBK clarified its communications by stating that ‘this upward adjustment of the CBR was expected to provide a signal to banks that interest rates should rise and therefore reduce the expansion in credit to the private sector’ Central Bank of Kenya (2011d). 26 Bank of Uganda (2011). 27 International Monetary Fund (2011b). 28 National Bank of Rwanda (2011a). 29 The average maturity of loans in the cleaned loan-level Uganda dataset of Abuka et al. (2015) is 1.5 years. Moreover, it is a characteristic of fully credible regimes that very long rates do not move much with short rates, because inflation expectations are well anchored (Gurkaynak et al., 2007). See also Bulir and Vlcek (2015). 30 While not necessarily relevant in this episode of dramatic tightening, this point is closely related to the argument, e.g., in Woodford (2001), that the substantial smoothing seen in advanced-country monetary policy reaction functions implies a large pass-through to lending rates. A lesser degree of smoothing thus implies lower pass-through. 31 They also identify a bank-balance-sheet channel in which balance sheet conditions of banks influence these effects, with for example poorly capitalised banks transmitting the interest rate changes more strongly. 32 Tanzania’s also relatively closed capital account might have been expected to mitigate the effects of monetary policy shocks on the exchange rate. In practice, the degree to which the nominal exchange rate is allowed to respond to interest rate differentials depended on the set of relatively slowly-moving policies on capital account openness, with a more closed account implying less reaction, and also on policies such as sterilised intervention and more temporary administrative measures that varied substantially through time and which are hard to measure directly. 33 In Uganda, the output gap increased from Q32010 through Q32011 despite positive real interest rates for the first half of that period. Fiscal policy was expansionary during the first part of this period, with a fiscal impulse (the change in the primary balance adjusted for the cyclical position, estimated at 1% of GDP for the Q3:2010-Q2:2011 period) (International Monetary Fund, 2011a). 34 Uganda stabilised somewhat earlier, in August, in this case coinciding more closely with the levelling off of the VIX but also with the earlier beginning of monetary policy tightening phase, which also peaked in October. 35 REER is downloaded from IMF database. The median and percentiles are calculated from REER of Brazil, Chile, Colombia, Czech Republic, Estonia, Ghana, Hungary, India, Indonesia, Israel, Korea, Mauritius, Mexico, Pakistan, Peru, Philippines, Poland, Romania, Serbia, Slovenia, South Africa, Sri Lanka, Thailand and Turkey. 36 OLS results without pre-filtering (available on request) tell a similar story, with if anything a smaller role for the exogenous variables. 37 That is, we create a ‘counterfactual’ based on the predictions of the estimated model given the exogenous variables, rather than using actual lagged values of the endogenous variable. 38 The negative coefficient on the US Rate in the exchange rate regression may be considered puzzling. A higher US rate should lead to a spot depreciation but a subsequent appreciation would be implied by uncovered interest parity. Because this is a regression using monthly data, either sign could be observed. 39 Mishra et al. (2014) find evidence to this effect in a large sample of developing countries. 40 Saxegaard (2006). 41 This evidence suggests that these ‘excess’ reserves may be an equilibrium phenomenon reflecting factors such as risk aversion on the part of banks and structural deficiencies in the functioning of the interbank market. Addressing these deficiencies that create excess liquidity is likely important for the promotion of financial development but, judging from this episode, it does not appear necessary for the transmission of monetary policy. References Abuka C. , Alinda R. , Minoiu C. , Peydro J. , Presbitero A. , 2015 , ‘Monetary Policy in a Developing Country: Loan Applications and Real Effects,’ Working Paper 15/270, International Monetary Fund, Washington, D.C. Adam C. , Maturu B. , Ndung’u N. , O’Connell S. ( 2010 ) ‘Building a Monetary Regime for the 21st Century’, in Adam C. , Collier P. , Ndung’u N. (eds) , Kenya: Policies for Prosperity . Oxford : Oxford University Press . Andrle M. A. , Berg A. , Berkes E. , Morales R. , Portillo R. , Vavra D. , Vleck J. , Money Targeting in a Modern Forecasting and Policy Analysis System: An Application to Kenya , Working Paper 13/239, Washington : International 2013 , Monetary Fund . Andrle M. A. , Berg A. , Morales R. , PortilloWo R. , Vleck J. , 2013 , Forecasting and Monetary Policy Analysis in Low Income Countries (1): Food and Non-Food Inflation in Kenya. Working Paper, International Monetary Fund, Washington, D.C. Bank of Uganda , 2011 , Monetary Policy Statement for October 2011. Berg A. , Charry L. , Portillo R. , Vlcek J. , 2013 , The Monetary Transmission Mechanism in the Tropics: A Narrative Approach. Working Paper, International Monetary Fund, Washington, D.C. Berg A. , O’Connell S. , Pattillo C. , Portillo R. , Unsal F. , 2015 , Monetary Policy Issues in Sub-Saharan Africa, The Oxford Handbook of Africa and Economics: Volume 2: Policies and Practices. Bulir A. , Vlcek J. , 2015 , ‘Monetary Transmission: Are Emerging Market and Low Income Countries Different,’ Working Paper 15/239, Washington: International Monetary Fund. Central Bank of Kenya , 2011 a, MPC Press Release. March. Central Bank of Kenya , 2011 b, MPC Press Release. July. Central Bank of Kenya , 2011 c, MPC Press Release. September. Central Bank of Kenya , 2011 d, MPC Press Release. October. Chinn M. , Ito H. ( 2008 ) ‘ A New Measure of Financial Openness ’, Journal of Comparative Policy Analysis , 10 ( 3 ): 309 – 22 . Google Scholar Crossref Search ADS Christiano L. J. , Eichenbaum M. , Evans C. L. ( 1999 ) ‘Monetary Policy Shocks: What Have We Learned and to What End?’, in Chapter 2 of Handbook of Macroeconomics , Vol 1 . London : Elsevier , pp. 65 – 148 . Part A. Cottarelli C. , Kourelis A. ( 1994 ) ‘ Financial Structure, Bank Lending Rates, and the Transmission Mechanis of Monetary Policy ’, Staff Papers—International Monetary Fund , 4 ( 4 ): 587 – 623 . Google Scholar Crossref Search ADS Davoodi H. , Dixit S. , Pinter G. , 2013 , Monetary Transmission Mechanism in the East African Community: An Empirical Investigation. Working Paper, International Monetary Fund, Washington, D.C. Dotsey M. , Reid M. ( 1992 ) ‘ Oil Shocks, Monetary Policy, and Economic Activity ’, Federal Reserve Bank of Richmond Economic Review , 78 : 14 – 27 . Friedman M. , Schwartz A. ( 1963 ) A Monetary History of the United States . Princeton : Princeton University Press , p. 1963 . Gurkaynak R. , Levin A. , Mardar A. , Swanson E. ( 2007 ) Inflation Targeting and the Anchoring of Inflation Expectations in the Western Hemisphere , Economic Review, San Francisco : Federal Reserve Bank of San Francisco . International Monetary Fund ( 2008 ) Regional Economic Outlook . Washington, DC : Sub-Saharan Africa . April. International Monetary Fund , 2011 a, ‘Uganda—Third Review Under the Policy Support Instrument, Request for Waiver of Nonobservance of an Assessment Criterion, and Request for Modification of Assessment Criteria.’ Washington, D.C. December 21. International Monetary Fund , 2011 b, ‘Tanzania: Letter of Intent, Memorandum of Economic and Financial Policies, and Technical Memorandum of Understanding,’ December 23. International Monetary Fund , 2012 , Annual Report on Exchange Arrangements and Exchange Restrictions. Washington, D.C. Retrieved 2011, from http://www.imfareaer.org/Areaer/Pages/Home.aspx International Monetary Fund , 2015 , Evolving Monetary Policy Frameworks in Low-Income and Other Developing Countries – Staff Report, Washington, D.C. October 23. Kasekende L. , Brownbridge M. ( 2011 ) ‘ Post-Crisis Monetary Policy Frameworks in Sub-Saharan Africa ’, African Development Review , 23 ( 2 ): 190 – 201 . Google Scholar Crossref Search ADS Li B. , O’Connell S. , Adam C. , Berg A. , Montiel P. , 2016 , VAR Meets DSGE: Uncovering the Monetary Transmission Mechanism in Low-Income Countries, Working Paper 16/90, International Monetary Fund, Washington D.C. Mishra P. , Montiel P. , 2013 , ‘How Effective is Monetary Transmission in Low Income Countries? A Survey of the Empirical Evidence’ , Economic Systems , 37 ( 2 ): 187 – 216 . Google Scholar Crossref Search ADS Mishra P. , Montiel P. , Pedroni P. , Spilimbergo A. ( 2014 ) ‘ Monetary Policy and Bank Lending Rates in Low-Income Countries: Heterogeneous Panel Estimates ’, Journal of Development Economics , 111 : 117 – 31 . Google Scholar Crossref Search ADS Mishra P. , Montiel P. , Spilimbergo A. ( 2012 ) ‘ Monetary Transmission in Low-Income Countries: Effectiveness and Policy Implications ’, IMF Economic Review , 60 : 270 – 302 . Google Scholar Crossref Search ADS Montiel P. , Adam C. , Mbowe W. , O’Connell S. , 2012 , Financial Architecture and the Monetary Transmission Mechanism in Tanzania. Working Paper, International Growth Center. National Bank of Rwanda , 2011 a, The BNR Key Repo Rate Increased From 6% to 6.5%. O’Connell S. , Pattillo C. , Portillo R. , Unsal D. , 2015 , Signaling and the Transmission of Monetary Policy: The Rold of the Policy Regime, Unpublished Manuscript (Washington: International Monetary Fund). Romer C. , Romer D. ( 1989 ) ‘Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Friedman’, in NBER Macroecononomics Annual , vol. 4 . Cambridge : National Bureau of Economic Research , pp. 121 – 84 . Saxegaard M. , 2006 , Excess Liquidity and Efectiveness of Monetary Policy: Evidence from Sub-Saharan Africa. Working Paper, International Monetary Fund. Shapiro M. D. ( 1994 ) ‘Federal Reserve Policy: Cause and Effect’, in Mankiw N. G. (ed.) , Monetary Policy . Chicago : University of Chicago and NBER . Stone M. , 2003 , “‘Inflation Targeting Lite”’, WP/03/12, Washington: International Monetary Fund Summers L. ( 1991 ) ‘ The Scientific Illusion in Empirical Macroeconomics ’, The Scandinavian Journal of Economics , 93 ( 2 ): 129 – 148 . Google Scholar Crossref Search ADS Woodford M. , 2001 , “‘Monetary Policy in the Information Economy”’ in Economic Policy for the Information Economy: Proceedings of the Federal Reserve Bank of Kansas City Annual Symposium Conference Jackson Hole, Wyoming, August 30–September 1: 297-370. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of African Economies Oxford University Press

The Monetary Transmission Mechanism in the Tropics: A Case Study Approach

Loading next page...
 
/lp/oxford-university-press/the-monetary-transmission-mechanism-in-the-tropics-a-case-study-yUswukaSBU

References (13)

Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com.
ISSN
0963-8024
eISSN
1464-3723
DOI
10.1093/jae/ejy022
Publisher site
See Article on Publisher Site

Abstract

Abstract Many central banks in low-income countries in Sub-Saharan Africa are modernising their monetary policy frameworks. Standard statistical procedures have had limited success in identifying the channels of monetary transmission in such countries. Here we take a case study approach and centre on a significant tightening of monetary policy that took place in 2011 in four members of the East African Community: Kenya, Uganda, Tanzania and Rwanda. We find evidence of the transmission mechanism in most of the countries. Variations across countries can be explained mainly by differences in the policy regime. Skillful exploitation of natural experiments that provide identifying variation in important variables represents the best hope for increasing our empirical understanding of macroeconomic fluctuations. While lacking the scientific pretension of an explicit probability model, careful historical discussions of events surrounding particular monetary changes, such as those provided by Friedman and Schwartz (1963), persuade precisely because they succeed in identifying relevant natural experiments, and describing their consequences. (Summers, 1991, p. 130) 1. Introduction Macroeconomic outcomes in many Low-Income Countries (LICs) in Sub-Saharan Africa (SSA) have greatly improved in recent years. Many central banks in the region are now looking to modernise their frameworks to make policy more forward looking, to further promote macroeconomic stability and ultimately growth and financial development. At the same time, uncertainty about how to proceed abounds. At the heart of this uncertainty is the question of how different LICs are in ways that matter for monetary policy. How does monetary policy transmit to the economy? An influential line of argument proposes that monetary transmission in LIC economies is weak and unreliable, greatly limiting the scope for monetary policy to serve as a key policy tool for macroeconomic stabilisation.1 Indeed, studies on the effects of monetary policy in activity and prices in LICs have largely relied on the use of statistical techniques such as VARs and often find weak or insignificant effects of monetary policy. This may reflect the nature of the underlying transmission mechanism. However, it may also reflect limitations of the VAR and related techniques in the LIC environment.2 The challenges to identifying the transmission mechanism in the data are big anywhere and probably larger in LICs. As elsewhere, policy is endogenous to events in the economy. More than elsewhere, the structure of the economy itself is evolving, for example as the financial system develops. Meanwhile, the transmission of policy itself depends on the policy framework, which is sometimes both obscure and changing. Finally, large supply shocks are frequent and data are noisy and scarce. Thus, analysis based on VARs—which require relatively long times series with consistent policy frameworks and plausible identification strategies—is unusually challenging.3 In this paper, we attempt, in the words of Summers (1991), to learn by ‘identifying relevant natural experiments and describing their consequences.’ In particular, we attempt to learn about the transmission mechanism by looking closely at a set of related cases in which policymakers in four countries (Kenya, Uganda, Tanzania and Rwanda, hereafter the EAC4) suddenly and unexpectedly tightened monetary policy to varying degrees. In two cases, at least, the policy adjustment was drastic. We thus apply a version of what Romer and Romer (1989) call the ‘narrative approach’ to identifying the effects of monetary policy: ‘The central element of this approach is the identification of monetary shocks through non-statistical procedures… The method involves using the historical record… to identify episodes when there were large shifts in monetary policy or in the behaviour of monetary policy that were not driven by developments on the real side of the economy.’4 To the best of our knowledge, this is the first attempt to apply a case-study methodology to monetary policy transmission in low-income countries. Some of the same features of the monetary policy environment in low-income countries that make VAR-based analyses extremely challenging also greatly complicate our approach too, however. In part because of the rapid evolution of monetary policy regimes, we are not able to identify several independent and comparable tightening events. And it is hard to argue, as do Romer and Romer (1989), that the tightening episodes are unrelated to recent economic developments, notably supply shocks.5 Thus, we rely mainly on a close reading of the narrative and the data to attempt to identify an unexpected component to the monetary policy tightening in question. We complement this narrative approach with two empirical exercises designed to help disentangle the independent role of monetary policy. First, we analyse high-frequency data on the supply shocks—to international commodity prices and global risk appetite—that are the main potentially confounding variables. This allows us to disentangle the timing of these supply shocks, changes in monetary policy, and movements in the exchange rate.6 Second, while we cannot statistically identify monetary policy shocks, we can take advantage of the exogeneity of the relevant supply shocks to identify the effects of these supply shocks per se on the exchange rate and inflation, conditional on the absence of monetary policy shocks. These exercises leave a significant role for monetary policy shocks themselves in explaining the dynamics of these variables, supporting our interpretation of the role of monetary policy shocks themselves. We are able to draw on variations across the four countries studies—notably in terms of economic structure and policy regime—to shed light on the influences of these factors on monetary policy transmission. Two ideas emphasised in the recent literature are salient. Montiel et al. (2012) argue that structural features of the LIC environment, notably underdeveloped and monopolistic financial systems and inflexible exchange rates, are likely to make transmission weak and unreliable, because policy rates do not transmit to lending rates in underdeveloped and monopolistic banking systems, and in any event these rates do not matter that much to the real economy. Other papers, e.g., Berg et al. (2015), and International Monetary Fund (2015), emphasise the role of policy regimes themselves in determining the nature of monetary transmission.7 These regimes in LICs tend to be complex and opaque, with a multiplicity of instruments (interest rates, quantities, foreign exchange intervention, etc.) and objectives (inflation, output, credit, exchange rates, etc.), and with the weights on both instruments and objectives often hard to discern and time-varying. This opacity has implications for transmission itself: where monetary aggregates are targeted, interest rate movements may not signal the policy stance clearly, for example.8 It also has implications for the analyst: even if a clear regime can be identified, it is likely to be of too-short duration to permit econometric analysis.9 Our narrative centres on a significant tightening of monetary policy that took place—to varying degrees and in different ways—in October 2011 in the EAC4. In 2010–2011, a major commodity price shock hit, and inflation took off in the EAC4, echoing the events of 2007–2008. Through 2010 and most of 2011, monetary policies remained fairly loose in Kenya, Uganda and Tanzania, with only cautious and ineffective efforts to tighten, perhaps encouraged by the experience of gradually moderating inflation without policy tightening during the earlier episode in 2009. The commodity price shocks turned out to be much more persistent this time, and they combined with vigorous economic activity, a negative balance of payments shock, and accommodative policy to further accelerate inflation and de-anchor expectations, weakening the exchange rate in an inflationary spiral. Some of the countries began to respond, to varying degrees. In July 2011, Uganda announced a new Inflation Targeting (IT) ‘lite’ policy regime and in August began a still-somewhat gradual tightening of policy. Kenya enacted fitful, partial, and ineffective tightening measures but did clarify its regime in September 2011. In Rwanda, in contrast, tighter monetary policy and a stable nominal exchange rate throughout the period kept inflation from taking off. Finally, the governors of the four central banks agreed at an unusual October 2011 meeting that policy needed to be tightened significantly in order to bring inflation under control, even at a cost to output, and they immediately acted.10 This tightening and surrounding events are our topic here. While the tightening took place in response to economic events, this does not make it entirely endogenous and thus invalidates our narrative approach to identifying the monetary transmission mechanism. Throughout 2011, concerns about the adequacy of the policy stance were increasingly widespread. However, it was unclear when a tightening might come or how strong it would be. Indeed, the narrative suggests that some observers were becoming concerned that it might not come at all. Thus, when it came, it was at least partly unexpected, unusual in the language of Friedman and Schwartz (1963). We can thus ask, what did this large monetary policy tightening shock do? We find some evidence consistent with a clear transmission mechanism. In some of the four countries, after a large policy-induced rise in the short-term interest rate, lending and other interest rates rose, the nominal and real exchange rate tended to appreciate, output tended to fall, and inflation declined. The variation in experiences among the four cases is informative. Most importantly, the cross-country variation in transmission seems to depend sharply on the policy regime in place. In particular, we find the clearest transmission in Uganda, where the IT lite regime itself was simpler and more transparent, and in Kenya, particularly once the authorities explicitly signalled the monetary policy stance with a short-term interest rate and described their intentions in terms of their inflation objective. In regimes where the stance of monetary policy was harder to asses such as Tanzania, which conducted monetary policy under a de jure monetary targeting regime, and Rwanda, a de facto exchange rate peg, the transmission of monetary policy to lending rates and, of course, the exchange rate is less evident. Nonetheless, in Tanzania, the exchange rate seemed to respond strongly to adjustments of the monetary policy stance. We see mixed signs of the importance of financial development. All four countries, like other LICs, have relatively small, concentrated, and bank-dependent financial systems, to varying degrees. In particular, Kenya’s large financial sector makes it an outlier, and it also had perhaps the most complete and unambiguous transmission. However, Uganda, which also clearly demonstrated the main elements of monetary policy transmission, has a relatively small financial sector comparable to that of the other two countries. While the shock was not entirely expected, it was not isolated from outside influences, particularly shocks to global risk appetite and commodity prices, which directly affected exchange rates and prices of traded goods. A close reading of the timing and some statistical evidence suggests that such shocks do not explain most of the exchange rate and price movements around the time of the tightening event, such that the residual unexplained component is consistent with our emphasis on the role of monetary policy itself. Moreover, Uganda’s somewhat earlier policy tightening—starting after its regime change in July—matches an earlier if also somewhat more gradual turnaround in its exchange rate and inflation. The paper proceeds as follows: We first briefly present the stylised facts of the countries under study, including structural features of the economy, the financial system, prices, and the policy regimes. We then proceed to the event study, identifying the policy shock and tracing out the effects of these shocks on the main macroeconomic variables: interest rates, credit aggregates, the exchange rate, output, and inflation. We then analyse more closely the role of important exogenous shocks to capital flows commodity prices that complicate interpretation of the events. Finally, we draw some tentative lessons. 2. The environment for monetary policy 2.1 Structure of the economy and the financial sector The four countries of interest here are in many ways typical of SSA LICs, making their experiences of general interest. Moreover, they share a recent macroeconomic history that is sufficiently stable that the recent monetary policy contraction is a salient event. They are among SSA’s many ‘success stories’ since the mid-1990s, with the achievement of macroeconomic and political stability and rapid economic growth (Table 1). Their economic structure is also broadly characteristic of SSA LICs: low-trade shares, mostly commodity exports, high though falling aid dependence, service-sector-led growth, and a large agricultural sector and rural population. The terms of trade have been fairly stable or rising in recent years.11 Table 1: Basic Economic Indicators, 2011 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Sources: IMF and the World Bank. Table 1: Basic Economic Indicators, 2011 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Country Population (Millions) Real GDP Per Capita (USD, PPP) Average Real GDP Growth (Percent, 2001–2011) Public Debt/GDP (Percent, 2011) Kenya 42 476 4.2 43.0 Uganda 35 374 7.3 23.6 Tanzania 46 460 6.9 27.8 Rwanda 11 360 7.9 23.1 Sources: IMF and the World Bank. Inflation in the EAC4 is volatile and highly correlated across countries, mostly explained by the high share of food items in the overall CPI. The weight of food prices in the CPI is highest in Tanzania (47%), followed by Kenya (36%), Rwanda (35%) and Uganda (27%).12 Domestic food crops yields in the region are highly dependent on weather patterns, which are characterised by a bimodal annual rainfall cycle. The existence of trade barriers to protect the domestic agricultural production has increased the sensitivity of the domestic food supply to unfavourable weather conditions. Financial systems also share the characteristics that have been associated with weak transmission, though with important cross-country variation. Financial development has been rapid in recent years, but these countries still generally have small and concentrated private-bank-dominated financial systems, a large informal financial sector, shallow capital markets (except Kenya), and short yield curves (except Kenya and more recently Uganda) (Table 2). In all four countries, commercial banks maintain substantial excess reserve deposits at the central bank. Additionally, the four economies exhibit different degrees of financial openness, with Uganda and Kenya being the more financially integrated of the four and Rwanda and Tanzania being the least. Looking across the four countries, Kenya stands out for its relatively developed financial sector, with a larger and less concentrated banking system. Table 2: Financial Sector Indicators, 2011 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Sources: IMF and World Bank. The first four indicators measure aspects of financial depth, while the last column measures financial openness. 1/ Assets of five largest banks as a share of total commercial banking assets. 2/ Index values are for 2010. The index measures the intensity controls over current or capital account transactions, the existence of multiple exchange rates, and the requirements of surrounding export proceeds. It takes a maximum value of 2.5 for the most financially open economies and a minimum of -1.9 for the least financially open. See Chinn and Ito (2008). Table 2: Financial Sector Indicators, 2011 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Groups Credit to the Private Sector (Percent of GDP) Bank Credit to the Private Sector (Percent of GDP) Five-Bank Asset Concentration (Percent) 1/ Stocks Traded, Total Value (Percent of GDP) Chinn-Ito Financial Openness Index 2/ Kenya 38.1 33.6 60.5 2.6 1.1 Uganda 17.9 13.8 73.6 0.1 2.5 Tanzania 17.8 15.8 67.6 0.1 −1.2 Rwanda 16.9 13.2 100.0 n.a. −0.9 Average EAC4 22.7 19.1 75.4 0.9 0.4 Low-Income Countries 19.6 18.8 80.0 4.9 −0.4 Emerging Economies 60.9 49.1 69.6 26.6 0.3 Advanced Economies 145.3 133.7 84.8 70.2 2.2 Sources: IMF and World Bank. The first four indicators measure aspects of financial depth, while the last column measures financial openness. 1/ Assets of five largest banks as a share of total commercial banking assets. 2/ Index values are for 2010. The index measures the intensity controls over current or capital account transactions, the existence of multiple exchange rates, and the requirements of surrounding export proceeds. It takes a maximum value of 2.5 for the most financially open economies and a minimum of -1.9 for the least financially open. See Chinn and Ito (2008). It is not straightforward to characterise the exchange rate and monetary policy regimes of these countries. All four countries reported having floating exchange rate regimes during this period. However, this paper follows the de facto classification in International Monetary Fund (2012), according to which Kenya, Uganda, and Tanzania had floating regimes (with Kenya having only limited foreign exchange intervention and hence labelled a ‘free float’ while the other two at times engaged in substantial intervention and were labelled ‘floats.’) Rwanda was classified as a ‘crawling-peg-like arrangement’ during the period in question (Appendix 1. Table A. 1 reports exchange rate, capital control, and monetary policy regimes for each country during the period in question). Meanwhile, Kenya, Uganda and Rwanda had de jure relatively open capital accounts insofar as there were relatively few official restrictions on capital account transactions, while Tanzania had widespread capital controls in place. As in most SSA countries with de jure floats, all four (except Uganda after July 2011, which labelled its regime ‘IT lite’ after that point) conducted monetary policy under a de jure monetary aggregate targeting framework, in principle adjusting the money supply to achieve intermediate targets in terms of broad money growth.13 However, these three regimes were much more flexible, and complex, in practice (Appendix 1. Table A.1). In particular, the experience of these countries matches the broader experience with such de jure money targeting regimes, which is that target misses are frequent, and at least in relatively low-inflation environments such deviations are not associated with misses of inflation objectives. Rather, central banks tend to make judgments on an ongoing basis as to whether targets should be achieved, and if not how they should be revised for the next quarter, depending on outcomes in money and exchange rate markets and a broader sense of whether inflation and output (and other) objectives are being achieved. This is for various reasons, not least because to adhere to the targets would generate excessive volatility of short-term interest rates in the face of money demand shocks. This makes the stance of policy hard to grasp and in particular very hard to infer from monetary aggregates themselves.14 Tanzania and especially Rwanda adhered most closely to their de jure money targeting regime, though even here economically meaningful deviations were frequent.15 Kenya had over time paid less and less attention to monetary aggregates, culminating in September 2011 with a clear announcement of a move to use the short-term interest rate as its main policy instrument, with the objective of achieving its inflation objectives (see Andrle et al. 2013). Uganda too had undergone an important evolution in this regard, moving in October 2009 from quite strict to flexible money targeting with substantial attention to interest rates as operating target to an explicit ‘inflation targeting ‘lite’ regime in July 2011 with use of short-term interest rates in pursuit of its inflation objectives. In sum, Kenya (especially after September 2011) and Uganda (especially after July 2011) adhered reasonably closely to an IT-style regime, with a managed float, a reasonably open capital account, interest rates as monetary policy instrument, and inflation stability as the clearly stated objective of monetary policy. Tanzania had a more complex regime, with a more heavily managed float, a multiplicity of monetary policy instruments and intermediate targets, and a less clear statement that inflation stability was the main objective of monetary policy. Finally, Rwanda seems to have largely subordinated its monetary policy to its de facto crawling peg exchange rate, though the de facto somewhat closed capital account may have given it some room for manoeuvre on monetary policy as well. The different degree of attention to monetary aggregates in practice has corresponded to varying degrees of interbank interest rate volatility. This volatility is clearest in Tanzania and least in Uganda and Kenya, consistent with their de facto use of interest rates as operating targets and indicators of the stance of policy. In Rwanda, lack of market clearing even in the interbank market seems to have allowed some degree of disconnect between money aggregates and short-term interest rates. In these hybrid regimes, it is often difficult to know how to interpret any particular interest rate. Sometimes ‘policy rates’ are not necessarily market clearing, present no arbitrage opportunities with other short-term interest rates, and may contain no signal of policy intention. But this can change suddenly as the details of central bank operations change. Pressures from fiscal authorities can encourage these central banks to create deviations between ‘policy rates’ and some market rates, particularly those at which the Treasury finances its activities, sometimes creating further opacity with respect to interest rates. In assessing the stance of policy, we in general refer to the interbank rate as an indicator of the market short-term interest rate, with due reference to ‘policy rates,’ exchange rate interventions, and money aggregates as appropriate. Other instruments, such as changes in reserve requirements, may have some independent signalling import but also are likely to work at least in part through their influence on interbank rates. 3. The event study We define a shock as an episode in which a central bank undertakes overt and unusual—large and substantially unexpected—actions to exert a contractionary influence on the economy in order to reduce inflation. The tightening we consider here took place mainly in October 2011 around the time of a meeting of EAC Central Bank governors at which it was stated that inflation was getting out of control and that monetary policy needed to be tightened. This meeting and the resulting sharp policy actions represent the distinct monetary policy shock that allows us to trace the transmission mechanism. We now take a closer look at the tightening episode. 3.1 The run-up Through mid-2011, international prices of food and energy shot up by more than 30% and 40%, respectively (Figure 1, Panel 1). Meanwhile, economic activity in the EAC4 was generally recovering from the earlier effects of the global financial crisis (Panel 2). Direct evidence suggests that the monetary policy stance was mostly accommodative, with nominal rates fairly flat and real interest rates mostly negative in all three countries (Panels 3 and 4). Partly reflecting this policy stance, but also at times owing to pressures on the capital account from swings in global risk aversion, nominal real exchange rates were generally weakening (Panels 5 and 6).16,17 Figure 1: View largeDownload slide The Run-Up. Sources: IMF estimates, Haver, national authorities. See footnote 19 for detailed calculation. Figure 1: View largeDownload slide The Run-Up. Sources: IMF estimates, Haver, national authorities. See footnote 19 for detailed calculation. By 2011Q3, these factors were reflected in headline inflation in Kenya, Uganda and Tanzania that surpassed the common inflation target of 5% (Panel 7).18 Even though most of the increase in headline inflation during this period is explained by the acceleration in food and fuel inflation, core inflation also increased substantially in all countries, almost doubling during the course of a year, and in all cases considerably overshooting the common inflation target (Panel 8).19 Consistent with the negative real interest rates, the monetary policy authorities were generally ‘behind the curve’ in responding to the building inflationary pressures.20 Moreover, policy responses were generally poorly signalled both in terms of the statements of the authorities and in that different instruments, such as different short-term interest rates, gave different signals. As inflation in Kenya continued to increase and the nominal exchange rate depreciated by about 10% in the period from March until September 2011, the Central Bank of Kenya (CBK) responded fitfully and opaquely. For example, a March increase of the Central Bank Rate (CBR) by 25 basis points to 6%, reversing a lowering of 25 basis points in January 2011, was accompanied by mixed signals as to the intent of the CBK, and had no discernible effect on lending rates or the exchange rate.21 In August 2011, the central bank resorted to stronger moves, including restricting access to the discount facility and restricting liquidity provision through open market operations, still keeping the CBR unchanged. These moves resulted in a brief intra-month 2,200 basis point spike in interbank rates, which had little apparent effect on treasury bill rates and none on lending rates or the exchange rate, which continued to depreciate. In September, the CBK clarified its operating regime, emphasising that it would use the CBR as its main policy instrument with the objective of achieving its inflation objectives. However, while the central bank raised the CBR from 6.25% to 7%, at the same time, and in contradiction, it provided liquidity to the interbank market at 5.75%. Meanwhile, policy statements from the CBK were ambiguous and lacked a clear announcement of policy tightening.22 Tanzania began acting as early as 2010Q4, by some measures. A sharp contraction in the real growth rate of money in late 2010 caused a jump in the interbank rate, with a hint of pass-through to T-bill rates, but this proved short lived even as the growth rate of money continued to slow. As real money growth reached nearly zero in mid-2011, interbank (but not T-bill) rates again spiked. There was no discernible effect of these actions on lending rates, the exchange rate, or credit. Uganda, in contrast, began effective tightening earlier. Interbank rates increased by some 500 basis points during the first half of 2011, though with little apparent effect; for example, lending rates remained unchanged. The BOU tightened much more consistently and coherently after the July 2011 announcement of a new ‘IT-lite’ regime and the introduction of the CBR to signal the stance of policy.23 The central bank raised the CBR in two steps from 13% in July to 16% in September. In this case, the lending rate increased by 340 basis points from July until September 2011, and the exchange rate began to stabilise, with the real exchange rate appreciating in September. We return to the effects of these measures in the next section. Rwanda presents a contrasting picture in terms of its policy regime and stance: through 2010–2011 output was below trend, the real exchange rate and real interest rates were stable (and the latter were positive). This may have reflected a tighter monetary policy stance, as well as the de facto crawling peg exchange rate regime. It is hard to infer much about the stance of policy during this period from the monetary aggregates; for example, real reserve money growth varied sharply in a way that seems unrelated to short-term interest rates and the exchange rate or for that matter inflation and the output gap. 4. The event and its aftermath As 2011 unfolded, inflation accelerated on the strength of higher food and oil inflation, strong demand, weakening exchange rates, and still-negative interest rates (Rwanda is the exception). The EAC4 came to the realisation by their October meeting that action was needed to stabilise the situation. On October 5th, the CBK increased the CBR by 400 basis points and on November 1 by a further 550 basis points to 16.5%. During this period, it also increased the cash reserve requirement by 75 basis points to 5.25%, and adjusted the discount window rate by more than 20%, ‘as decisive and immediate action is required from the monetary policy side to stem these inflationary expectations.’24 In addition, the Ministry of Finance lowered the foreign exchange exposure limit for commercial banks to 10% of core capital from 20%.25 The Bank of Uganda (BUG) followed up early substantial tightening, and the introduction of its new regime in July, by raising its policy interest rate by 400 basis points in October and a further 300 basis points in November to 23% and stepped up its intervention in the foreign exchange market to contain the depreciating pressures on the Shilling, stating that ‘the upside risks to inflation have increased, it is necessary to tighten monetary policy further… (this) should be seen as a clear signal of the BOU’s determination to bring inflation under control. However, should the upside risk to inflation continue in the months ahead, then monetary policy will be tightened further.’26 In Tanzania, the central bank increased its policy rate by 200 basis points in October and a further 202 basis points in November to 11.00%, and on October 26 augmented the minimum reserve requirements on government deposits held by banks from 20% to 30%, reduced commercial banks’ limit on foreign currency net open positions from 20% to 10% of core capital, tightened capital controls, and increased sales of foreign exchange in the interbank market.27 This shift was more decisive than earlier efforts, perhaps partly because this time the authorities put more emphasis on the policy rate, as well as on the quantitative actions. Finally, the National Bank of Rwanda took a variety of much more moderate tightening measures, consistent with the much lower degree of disequilibrium throughout 2011. Again, it is hard to point to any specific measurable action with respect to money aggregates, but the central bank’s policy rate was increased by 50 basis points to 6.5%, as ‘the Central Bank finds it appropriate to review its policy rate in order to keep the monetary aggregates at optimal levels to limit inflation pressures while continuing to support economic growth.’28 Having identified the policy-tightening event and the variations across the four countries, we now turn to an assessment of the various channels of transmission of monetary policy across our group of countries. The variety of instruments and (time-varying) differences in regimes can make it difficult to characterise in a simple way the stance of policy itself. We generally use the interbank rate as the best single measure of the policy shock itself. As a measure of the short-term market rate in all four countries, it captures aggregates the effects of policy rates and quantitative policies (quantitative interventions, reserve requirements) and distills the divergent effects of potentially inconsistent actions taken with various not-necessarily-market-clearing ‘policy rates.’ 4.1 The interest rate channel In Uganda and Kenya, the pass-through from policy rates to interbank rates was fast and complete (Figure 2, Panels 1 and 2) (and to T-bill rates, shown in the working paper). Tanzania’s battery of measures also transmitted quickly to money market rates. Figure 2: View largeDownload slide The Monetary Policy Contraction and its Aftermath. Source: IMF estimates, Haver, national authorities. Figure 2: View largeDownload slide The Monetary Policy Contraction and its Aftermath. Source: IMF estimates, Haver, national authorities. The relevance of the policy regime is evident in the transmission to banking rates (Panel 3). Lending rates, in particular, responded swiftly—though partially—to the monetary policy contraction in Kenya and Uganda. Uganda’s lending rates began to respond somewhat earlier, corresponding to the August/September tightening. There is little sign of transmission to lending rates in Tanzania and Rwanda. The lack of response of lending rates in Tanzania and Rwanda, despite the increases in the interbank rates in these two countries, is reminiscent of the non-response of lending rates to Kenya’s August spike in interbank rates. Even in Kenya and Uganda, the pass-through from short-term market to lending rates was partial. This may reflect lack of competition or other structural weaknesses in the financial system, as argued in Montiel et al. (2012). However, lending rates are longer-term rates, so partial pass-through to a (at least somewhat temporary) tightening is also consistent with fully functioning markets, by which we mean arbitrage across returns of different assets, and an expectation that the period of high short rates will be somewhat shorter than the tenor of the loans.29 Moreover, standard data in these countries (such as we use here) report average, not marginal lending rates. Finally, as we have already argued above, the pass-through of policy rates to short-term market and lending rates will depend on the clarity of the regime and in particular the ability of market participants to infer that the increase reflects policy intent and will not be quickly reversed.30 It may be that there was still some uncertainty about whether the authorities in Kenya and Uganda would stay the course. 4.2 The bank lending channel The data indicate the existence of the credit channel in Kenya, Uganda, and Tanzania. In these three countries growth in credit to the private sector peaked soon after the policy contraction started and decelerated substantially as the monetary authorities stepped up the pace of tightening. Accordingly, during the 2011Q3 to 2012Q3 period, credit to the private sector growth in Kenya, Uganda and Tanzania decelerated. Again, Uganda’s contraction began a month or two before the others. There are also signs of credit rationing in the case of Tanzania: even though lending rates did not respond to the tightening, there was a meaningful impact on the quantity of credit extended to the economy. In Rwanda, there is little sign of slower credit growth, perhaps reflecting the much less significant tightening. Abuka et al. (2015) provide important supportive evidence for the bank lending channel in the case of Uganda. A unique data set of loan-level data spanning the period in question allows them to control for demand effects through region-industry dummies and aggregate variables such as GDP, directly. This allows them to interpret the effects of the policy shock as causal for credit supply and as not due to demand effects. They find that higher short-term market rates interest rates are associated with an increase in banks’ lending rates and reductions in loan volume at the extensive and intensive margins. The strength of this bank lending channel is significant, albeit about half of that observed in advanced economies studied with similar data and techniques.31 4.3 The exchange rate channel In Kenya, Uganda, and Tanzania, the increase in short-term interest rates was associated with a contemporaneous appreciation of the currency. Notably, this took place during the decisive tightening phase in 2011Q4, but not earlier when policy was more cautious and less clearly signalled. Uganda is again a partial exception insofar the exchange rate stabilisation began two months earlier, corresponding to its earlier tightening phase. Rwanda’s de facto pegged regime, relatively closed capital account, and more moderate tightening are apparent in the absence of large exchange rate movements.32 4.4 Output The tightening episode is associated with a contraction of output in Uganda and to a lesser extent in Tanzania (Panel 7). The absence of a visible decline in the output gap in Kenya is notable. Of course, factors other than monetary policy such as fiscal policy and foreign demand also influence the output gap, and it is difficult to measure the output gap in the context of frequent supply shocks.33 The Abuka et al. (2015) analysis based on loan-level data provides supportive analysis of the effects of this particular monetary policy shock on output in Uganda. By identifying differential loan supply effects in districts with varying banking sector conditions, and measuring real effects through night-time light output measured from satellites, they can plausibly identify the real effects of monetary policy acting through the bank balance sheet channel. They find that output does indeed contract more in those districts where banks have balance sheets, suggesting strong balance sheet channel for the monetary contraction. 4.5 Inflation Finally, the inflation rate came down sharply with the monetary contraction. Headline inflation began turning around rather quickly, within a month or two, presumably reflecting the rapid pass-through of exchange rate movements. The turnaround was sharpest in Uganda and Kenya but was apparent in Tanzania as well. Again, Rwanda shows a much more gradual pattern, reflecting the fact that inflation was never far from target. Core inflation followed, albeit much more gradually (Panel 9). 5. High-frequency and statistical evidence on the independent role of monetary policy As the above narrative suggests, the monetary policy shock of late 2011 was not entirely expected, but it was also not isolated from outside influences, particularly shocks to global risk appetite and commodity prices. Because these shocks directly affected exchange rates and prices of traded goods, as well as monetary policy, they may have led us to overemphasise the importance of monetary policy shocks per se. In particular, an alternative story emphasises that inflationary pressures were not a reflection of vigorous economic activity and loose monetary policy but a result of the higher food and fuel prices as well as global risk premium shocks that depreciated exchange rates. And along the same lines, the eventual reductions in inflation reflect the fading of the effects of these global shocks, rather than the monetary policy response. In this section, we take two approaches to desentangling these factors. First, we take advantage of the availability of high-frequency (daily) data on global risk shocks and on key associated endogenous variables (the exchange rates and domestic interest rates) to look closely at the timing of these shocks. In particular, we examine whether monetary policy shocks that are not coincident with risk premium shocks are correlated with movements in the exchange rate, and whether the sign of this correlation is consistent with the narrative above or rather with reverse correlation from the exchange rate movements to interest rates. To summarise, we find that increases in short-term domestic interest rates around the times associated with the monetary policy tightening are correlated with appreciation of the exchange rate and generally not with coincident improvements in global risk appetite, providing direct evidence in favour of the above narrative. Second, we. large movements in the exchange rate are associated with risk premium shocks or with can be associated with monetary policy shocks per se (as measured by the interest rate) or w. We also look at This evidence suggests that such shocks do not explain most of the exchange rate and price movements around the time of the tightening event, such that the residual unexplained component is consistent with our emphasis on the role of monetary policy itself. Moreover, Uganda’s somewhat earlier policy tightening—starting after its regime change in July—matches an earlier if also somewhat more gradual turnaround in its exchange rate and inflation. 5.1 A high-frequency analysis of the role of shocks to role of global risk appetite Are shifts in capital flows and global risk aversion able to provide an alternative explanation of the exchange rate dynamics during the run-up and following the coordinated tightening? The year 2011 was one of increased global risk aversion, with the rising political tensions in the Middle East associated with the Arab Spring, the sovereign debt crisis in Europe, and the downgrading of the credit rating of major industrial economies. This surely contributed to exchange rate pressures on the EAC4, but cannot plausibly explain the timing and magnitude of the real depreciations observed. The currencies of Kenya, Uganda, and Tanzania started to weaken in 2010, well in advance of the episode, standing during 2010 and 2011 amongst the most depreciated currencies in the emerging and frontier markets world (excluding fixed exchange rate regimes, Figure 3). Figure 3: View largeDownload slide Emerging and Frontier Markets Real Exchange Rates. (Real effective exchange rate based on CPIs, Jan. 2010 = 100, increase means appreciation35). Sources: IMF INS database and authors’ computation. Figure 3: View largeDownload slide Emerging and Frontier Markets Real Exchange Rates. (Real effective exchange rate based on CPIs, Jan. 2010 = 100, increase means appreciation35). Sources: IMF INS database and authors’ computation. Taking a closer look at the high-frequency data, swings in global risk appetite can partly explain the sharp depreciation during the July–September period and perhaps some of the appreciation that followed the monetary policy tightening. However, the timing of the turnaround indicates a strong independent role for the monetary policy contraction. Global risk appetite, as proxied by the VIX Index, deteriorated markedly in August and September as the credit ratings of the United States, Japan, and Italy were downgraded and Europe’s debt crisis intensified (Figure 4). Tensions in international capital markets eased by late September. However, the currencies of Kenya and Tanzania reached their lowest levels weeks later.34 The Kenyan shilling strengthened immediately on the second day of the policy tightening announcement, after staying near low levels despite improved global sentiment. The Tanzanian shilling, on the other hand, continued to weaken even after the announcement of the coordinated policy, interrupting its slide only after a further battery of measures was announced on the week of October 24. Figure 4: View largeDownload slide VIX (Index) and EAC4 Exchange Rates. Source: Bloomberg. Note: Exchange rates are local currency units/dollar; an increase indicates a nominal depreciation. Figure 4: View largeDownload slide VIX (Index) and EAC4 Exchange Rates. Source: Bloomberg. Note: Exchange rates are local currency units/dollar; an increase indicates a nominal depreciation. This evidence suggests a role for monetary policy per se in driving the exchange rate, in that the exchange rate appreciations coincided with the monetary policy tightening much more closely than with the improvements in international risk sentiment, when we look closely. 5.2 The role of global shocks: a statistical counterfactual In this section, we construct a statistical counterfactual to examine directly the view that global shocks to financial conditions and commodity prices are sufficient to explain the rise and subsequent fall in inflation and other main macroeconomic dynamics around the events in question. While we do not think there are enough data and stable enough regimes to identify statistically the effects of monetary policy through this episode, we can estimate the effects of these global (and thus exogenous) shocks on inflation and other key variables in the countries of interest here. By omitting monetary policy from the regressions, we implicitly attribute any effect of the coincident monetary policy response to these global shocks to the global shocks themselves. Thus, the fitted values from these regressions represent a counterfactual that assumes no explicit or unexpected monetary policy response to the global shocks, or that such a response has no independent effect on macroeconomic variables. Before we carry out this exercise, two points are worth noting. First, as shown earlier, core inflation (mainly excluding food and energy prices) also increased substantially through 2011, almost doubling and overshooting the target, before beginning to come back down, Second, movements in food and fuel inflation are themselves influenced by the monetary policy stance. Monetary policy influences the domestic price of imported food and fuel through the exchange rate. For locally produced (and not-fully-traded) food, monetary policy can work through aggregate demand. Thus, disentangling the contribution of monetary policy from that of supply shocks cannot solely be based on separating food/fuel and core inflation. Andrle et al. (2013) discuss these issues and apply a simple structural model to events in Kenya during 2011 and find that monetary policy accounts for much of the inflation dynamics, including the behaviour of domestic food prices. We estimate two simple regression models, explaining actual nominal exchange rate depreciation and inflation with only clearly exogenous variables. The identification strategy is to give the maximum possible influence to these variables, by interpreting any comovement with domestic policy variables such as the interest rate as being due to these exogenous variables. For exchange rate depreciation, the exogenous are the percent changes in the VIX and the Fed funds rate. For CPI inflation, the exogenous variables are US inflation and world food and oil prices. We assume an AR(2) process as follows: Δyt=c0+c1Δyt−1+c2Δyt−2+cXt+εt Where Δy is the endogenous variable (nominal exchange rate depreciation or inflation) and Xt is a matrix of exogenous explanatory variables. We use monthly data on nominal exchange rate depreciation and inflation in Kenya, Rwanda, Tanzania and Uganda. Both nominal exchange rate depreciation and inflation are month-on-month annualised changes, with the nominal exchange measured in local currency units per US Dollar. Spectral regression is used to estimate model parameters. This is equivalent to pre-filtering the data to include only selected frequencies and then applying OLS to this filtered data. This enables us to infer model parameters from the business cycle dynamics in the data while mitigating adverse effects of very high-frequency noise. This regression approach provides more robust parameter estimates, e.g., parameters remain relatively robust to changes in the data range. Furthermore, compared to standard OLS the spectral regression boosts effects of exogenous variables and increases the explanatory power of the regression, as measured by R2 statistics. Estimation outcomes are presented in Tables 3 and 4 and are inferred from cycles with periodicity of 4–96 months.36 Table 3: Regression Results—Nominal Exchange Rate Depreciation Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Source: Authors’ computation. The ‘US Rate’ is the Federal Funds rate. Note: Statistically significant parameters are denoted by stars (***, **, and * denote 99%, 95% and 90% confidence level). The ‘US Rate is the Fed Funds rate. The exchange rate is the nominal exchange rate measured in local currency units/US Dollar. Table 3: Regression Results—Nominal Exchange Rate Depreciation Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Kenya Rwanda Tanzania Uganda Constant 7.88*** 1.01 7.30*** 6.61*** Lagged (-1) 0.99*** 1.27*** 1.12*** 1.19*** Lagged (-2) -0.63*** −0.57*** −0.67*** −0.67*** US Rate -1.47** 0.09 −1.28** −1.15* Delta VIX 0.03*** 0.00 0.01* 0.03*** R2 0.48 0.46 0.49 0.56 Source: Authors’ computation. The ‘US Rate’ is the Federal Funds rate. Note: Statistically significant parameters are denoted by stars (***, **, and * denote 99%, 95% and 90% confidence level). The ‘US Rate is the Fed Funds rate. The exchange rate is the nominal exchange rate measured in local currency units/US Dollar. Table 4: Regression Results—Inflation Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Source: Authors’ computation. Note: Statistically significant parameters are denoted by stars (***, ** and * denote 99%, 95% and 90% confidence level). Table 4: Regression Results—Inflation Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Kenya Rwanda Tanzania Uganda Constant 3.7498*** 1.54** 0.02*** 1.20*** Lagged (-1) 1.2802*** 1.09*** 1.28*** −0.58*** Lagged (-2) −0.764*** −0.62*** −0.59*** 0.07*** US Inflation 0.0375* 0.24*** 0.03*** −0.03*** World Food Prices −0.018 −0.01 −0.01** 0.02** World Oil Prices 0.0057 0.00 0.00 1.92** R2 0.57 0.40 0.56 0.51 Source: Authors’ computation. Note: Statistically significant parameters are denoted by stars (***, ** and * denote 99%, 95% and 90% confidence level). We then construct counterfactual measures using only the exogenous variables along with the associate endogenous dynamics. The counterfactual measures represent what would be dynamics of endogenous variables without monetary policy shocks. The counterfactual measures are generated starting in 2011M04. The regression explains a fair amount of the variance of the exchange rate (from 56% in Uganda to 46% in Rwanda, presumably low owing to its quasi-managed regime). Most of this explanatory power is due to the importance of the lagged exchange rate itself, though the exogenous variables are highly significant. In Figure 5, we plot the predicted value of the exchange rate based only on the exogenous shocks and associated endogenous dynamics (the ‘counterfactual,’ which we interpret as reflecting the dynamics of the exchange rate in the absence of unexpected monetary policy decisions).37 As Figure 5 shows, in all but Rwanda the exogenous factors go in the right direction in explaining the depreciations in Q3 and subsequent appreciations, but much less than observed and with not quite the right timing. The peak depreciation of the predicted exchange rate in all countries occurs around September, but the actually (much weaker) bottom occurs in October in Tanzania and Kenya and in August in Uganda, more consistent with the timing of the monetary policy shock.38 Figure 5: View largeDownload slide Nominal Exchange Rate Depreciation, MoM Annualised in Percent. Source: Authors’ computation; see Table 3. Figure 5: View largeDownload slide Nominal Exchange Rate Depreciation, MoM Annualised in Percent. Source: Authors’ computation; see Table 3. The results for headline inflation paint a similar story. Inflation peaks coincide or follow with a one-month lag in the nominal exchange rate depreciation and they cannot be explained by the counterfactual (Figure 6). Figure 6: View largeDownload slide CPI Inflation, MoM Annualised in Percent. Source: Authors’ computation; see Table 4. Figure 6: View largeDownload slide CPI Inflation, MoM Annualised in Percent. Source: Authors’ computation; see Table 4. To summarise this section, the swings in global risk aversion and the food and fuel supply shock during 2011 did play an important role during the episode under study, they are only part of the story, both in terms of magnitude and timing, and do not overturn the conclusion that monetary policy seems to have played a decisive role. 6. Summary and interpretation We have identified a moment when three of the EAC4 broke from previous behaviour and executed more-or-less clearly signalled monetary policy contractions with the explicit intent of reducing inflation. We find clear evidence of most elements of the standard transmission mechanism in most of the countries. The transmission was the clearest in Kenya and Uganda, where market and lending rates followed the policy rate with little lag, the nominal exchange rate appreciated sharply on the policy announcement, credit growth (and, at least in Uganda, the output gap) began to decline immediately. Both headline and core inflation also began to decline almost immediately. Transmission was less clear in Tanzania, where the effects on some interest rates, activity, the exchange rate, and inflation are still broadly evident, but lending rates failed to respond and the effects on output are barely evident. Rwanda presents a control along several dimensions: initial imbalances were much smaller, the tightening much less significant, and the various components of transmission much more muted or invisible. Based on and summarising the preceding narrative, we can now evaluate the two hypotheses discussed in the instruction to explain the variation in cross-country experience. In the episode under study, we find substantial explanatory power in the idea that the nature of the policy regime conditions transmission. In the cases of clearest transmission, Kenya and Uganda, the regimes in October 2011 most resembled inflation targeting in that the authorities prioritised inflation, emphasised the role of the policy rate, allowed the nominal exchange rate a large degree of flexibility, and broadly avoided multiple objectives. Earlier tightening efforts by Kenya, e.g., in August 2011, were more incoherent in terms of the consistency across different instruments and communications and did not translate into lending rates or the exchange rates. Uganda’s earlier tightening efforts were more coherent and stronger following its July 2011 move to ‘IT lite’ in July and indeed had some effect on lending rates, credit, the exchange rate and inflation about two months before Kenya and Tanzania. In Tanzania instead, the money targeting regime led to highly volatile short-term interest rates, a variety of instruments were used in not-always-consistent ways, and overall there was less clear signalling of the policy stance. Rwanda’s regime was in some ways the most complex, with a quasi-pegged exchange rate, direct influence on private sector credit, direct influence on private sector credit, monetary aggregate targets and a policy rate. The emphasis on the exchange rate left room for monetary policy itself to act, and insofar as it could, the regime did not provide a clear signal. In the event, there was apparently less tightening, and less need to tighten. The second hypothesis is that transmission worked better in countries with greater financial depth and more open capital accounts. We find mixed support for this story in this episode. It remains plausibly the case that countries with more liquid and deeper financial markets will observe stronger transmission from policy rates to the macroeconomy.39 However, in this particular case measures of financial depth do not seem determinative for the clarity of transmission. A glance back at Table 2 reminds us that Kenya is the clear outlier for all measures of financial depth, with the other three countries remarkably similar. And yet, as we have argued, the evidence for transmission looks fairly strong, and similar, for Uganda and Kenya, in contrast to Tanzania and Rwanda. On the other hand, the narrative is consistent with the view that the lower degree of capital account openness in Tanzania and Rwanda (Table 2) may have contributed to obscuring or impairing transmission in these two countries. The exchange rate did seem to respond in Tanzania, but less dramatically than in Kenya and Uganda. Finally, it is often asserted that the high levels of excess reserves usually observed in SSA countries prevent the operation of the monetary transmission mechanism, for example because tightening policy may amount to ‘pushing on a string’ as banks respond to a contraction by withdrawing excess reserves.40 In the episode we examine here, excess reserves did not seem to impair the transmission mechanism. As Figure 7 shows, there are indeed substantial reserves in excess of required levels in all four countries, with large variations across time and countries but with no evident influence on transmission. While in Kenya, excess reserves fell during 2011, they remained above 6% of required reserves even in October, and they rose in Uganda and varied around 20% of required reserves during the peak of the tightening phase, higher than they had been since 2007.41 Figure 7: View largeDownload slide Excess Reserves. Source: Central Bank Data. Figure 7: View largeDownload slide Excess Reserves. Source: Central Bank Data. 7. Conclusions The identification of monetary policy transmission is difficult under any circumstances, and especially so in countries with poor data, obscure and time-varying policy regimes, and frequent supply shocks. As emphasised by Summers (1991) among others, the analysis of dramatic events such as the Great Depression and the Volker disinflation in the United States has played a critical role in forming professional opinion and framing the discussion in advanced countries. However, such analyses are scarce in developing countries. We have taken advantage of a dramatic tightening of monetary policy in four countries in East Africa in October 2011 to trace through the effects of this tightening on interest rates, credit, the exchange rate, output, and inflation. We find clear evidence of a working transmission mechanism in two of the countries: after a large policy-induced rise in the short-term interest rate in Kenya and Uganda, lending rates rose, the exchange rate appreciated, credit growth slowed, output growth tended to fall, and inflation declined. The other two countries represent a contrast to varying degrees. In Tanzania, some but not all market rates and the exchange rate seems to respond to adjustments of the monetary policy stance, credit growth slowed, and we see some signs of the effects of policy in output (possibly through credit rationing); in Rwanda, the initial disequilibrium was much smaller, the exchange rate more-or-less controlled, any monetary policy tightening much less evident, and any effects of such a policy tightening as took place therefore difficult to observe. These case studies provide many illustrations of the role that the policy framework itself plays in governing the strength of transmission. Most importantly, Kenya and Uganda by October had clarified their inflation objective and the centrality of the policy rate as the main signal. In this context, they were able to clearly articulate that they were raising rates to bring inflation back down. Earlier efforts in Kenya in August, before this clarification, were ineffective. Tanzania represents an intermediate case, in which a continued focus on money targeting and some inconsistency across policy instruments coexisted with a somewhat less clear transmission. The proliferation of policy instruments was common across all four countries. This put a premium on coherence and communication in signalling the policy stance. The difficulty in interpreting the different measures of policy may impact not only our ability as researchers to discern events but the capacity of interest rates to clear markets and signal policy. From a methodological point of view, this active use of a wide set of instruments under differing policy regimes with multiple objectives, along with the suggested quasi-contemporaneous nature of the transmission mechanism suggest that extra care should applied used when using standard statistical procedures, such as VARs, to measure the effects of policy, especially when those studies are conducted for country groups. In contrast, we found some role for financial openness but little for financial depth or the degree of excess reserves in explaining the cross-country patterns of transmission in these particular cases. We should not overemphasise or over-generalise this result. It remains plausible that countries with less developed financial markets and less open capital accounts will observe weaker transmission. However, this does not seem to have been a determinative feature here. Clearly, in general, shocks other than those to monetary policy are the main drivers of macroeconomic outcomes in countries such as those we examine here. And here as elsewhere, monetary policy was not made in a vacuum, and identification of shocks and their effects is challenging. It is always a great leap of faith to suppose that a residual in an estimated monetary policy reaction represents such a shock and not a misspecification (e.g., an omitted variable or a nonlinearity) in the reaction function. In our cases, we have used the historical narrative to attempt to argue that much of the shock was a surprise, and we have tried to exclude simple alternative hypotheses about the drivers of some of the key variables. But ultimately the results cannot be definitive. Much remains unknown about the transmission mechanism in these countries. The role of supply shocks, food prices, the banking system and limited financial participation, fiscal policy, limited capital account openness, and many other features deserve further exploration. The transmission mechanism appears to be working, particularly when signals are clear and the regime simple and coherent. Acknowledgements This paper is part of a research project on macroeconomic policy in low-income countries supported by the U.K.’s Department for International Development. The project draws in part on earlier work with IMF colleagues, particularly Rogelio Morales. We would also like to thank Chris Adam, Steve O’Connell, and seminar participants at the IMF’s African Department and the Center for the Study of African Economies in Oxford. All errors remain ours. Footnotes 1 See Mishra et al. (2012). 2 Li et al. (2016). For reviews of the literature, see International Monetary Fund (2008), Mishra and Montiel (2013), and Davoodi, Dixit and Pinter (2013) for the East African Community. 3 Li et al. (2016). 4 Romer and Romer (1989), pp. 1. 5 The exogeneity of Romer and Romer’s policy events is challenged in Shapiro (1994) and Dotsey and Reid (1992). See the discussion in Christiano et al. (1999). 6 We thank an anonymous referee for suggesting this interpretation of these data. 7 By ‘regime’ we mean the monetary policy and exchange rate regime, which are of course related. Textbook classifications involve pegs and floats, and money or inflation targeting. As described in more detail below and in Appendix I, the reality in these countries is substantially more complex. We include a discussion of capital mobility in characterising the policy regime, since it also conditions the role of monetary policy. 8 See O’Connell et al. (2015). Cottarelli and Kourelis (1994) show that low pass-through to lending rates can be attributed to the volatility of money market rates, which limits their information content regarding the stance of monetary policy and which is characteristic of the quantity-based monetary frameworks in place in these two countries. The presence of a high level of noise in money market rates can also hinder the degree of money market development, another structural feature identified in the literature as relevant in explaining the transmission of monetary policy impulses to lending rates. 9 Of course, these two sets of hypotheses need not be mutually exclusive. Montiel et al. (2012) emphasises both structural (financial underdevelopment) and regime-related (de facto pegged exchange rates) reasons for attenuated transmission. 10 In the communique from the October 12, 2011 meeting (available at ‘https://www.bou.or.ug/bou/bou-downloads/press_releases/2011/Oct/EAC_GOVERNORS_MEETING_ON_THE_CURRENT_CRISIS_-_OCTOBER_12_2011.pdf’), the ‘Governors observed that the region is facing very high inflation originating primarily from high food and fuel prices but also from demand pressures. The Governors also observed that the region is facing pressures on the currencies to weaken and exchange rate volatility. The pressures for the currencies to weaken result mainly from the widening of the current account deficit originating from rapid expansion of the oil import bill and imports for infrastructure development. In addition, the exchange rate volatility has been due to the effects of the Euro sovereign debt crisis and currency speculation activities. Given these challenges, the Governors agreed to coordinate the following actions: Tightening monetary policy, Stemming volatility in the foreign exchange markets, and Curbing currency speculation activities.’ The motivations for acting jointly presumably included the benefits of political solidarity in the context of the drive for economic union in the East Africa Community, a fear that disjoint action might result in undesired exchange rate fluctuations within the group, and a view that common action might have a stronger signalling effect. 11 Berg et al. (2013) presents much more background information and references. International Monetary Fund (2015) surveys the monetary policy environment, regimes and policy challenges in a broader context. 12 The Tanzania CPI survey includes rural households, unlike the surveys in Kenya, Uganda, and Rwanda. 13 ‘IT lite’ is a term coined (as far as we are aware) in Stone (2003), who characterise it as a regime in which countries that float their exchange rate and announce an inflation objective but are not able to maintain the inflation target as the foremost policy objective. It has come to be used to describe countries that adhere to something like ‘inflation targeting (IT)’ but without all the elements often considered to characterise such a regime. 14 See International Monetary Fund (2015),Appendix 2, Adam et al. (2010), and Kasekende and Brownbrige (2011) for discussions of this issue and Andrle et al. (2013) for Kenya specifically. 15 As shown in International Monetary Fund (2015), deviations that occur roughly once per year correspond in magnitude, by back-of-the-envelope calculations, to interest rate deviation of 5 percentage points in Rwanda and 20 percentage points in Tanzania. Rwanda’s close control of the nominal exchange rate created some tensions with the money targets in practice. 16 Section IV.G examines the role of exogenous external shocks more closely to disentangle them from that of policy. 17 Rwanda was an exception to these generalisations, with output below potential (though rising), a stable nominal exchange rate, and positive real interest rates. 18 Output gaps are estimated with a Hodrick-Prescott filter on the 4-quarter cumulative real GDP in Uganda and Tanzania and Non-Agricultural GDP in Kenya and Rwanda. The estimation sample includes data up to 2012Q3 to correct for end-of-sample bias. Real interest rates are calculated using the 12-month backward-moving average CPI-based inflation rate. The nominal exchange rates are bilateral with the US dollar. The real exchange rates are CPI-based bilateral rates with the US dollar. 19 To assess non-commodity inflationary pressures in a comparable way across the EAC4, we constructed an indicator of core inflation excluding food and fuel prices. These measures are not necessarily the same as those compiled by local authorities. For details, see Berg et al. (2013). 20 In part, this may have reflected the experience of the earlier episode of external price shocks in 2007/2008, when temporarily rising inflation was followed by the commodity price and external demand collapse of the global financial crisis, obviating the need for a monetary policy response in that case, as discussed in Berg et al. (2013). 21 In March 2011, the CBK suggested somewhat contradictorily that ‘this tightening will provide a solution to inflationary pressure and will stabilize the exchange rate while still protecting economic activity’ (Central Bank of Kenya (2011a)). In July, they still considered this action sufficient to mitigate soaring inflation (Central Bank of Kenya (2011b)). 22 At an extraordinary meeting on September 14, 2011, the CBK announced that ‘The high overall inflation environment is mainly a consequence of high food prices and high fuel and energy prices….The CBK will pursue the inflation objective through a continuation of the gradual tightening of monetary conditions.’ (Central Bank of Kenya (2011c)). 23 With the introduction of the IT-lite framework also came the release of a Monetary Policy Statement, signed by the governor and providing forward guidance to market participants. 24 Central Bank of Kenya (2011d). 25 In October 2011, the CBK clarified its communications by stating that ‘this upward adjustment of the CBR was expected to provide a signal to banks that interest rates should rise and therefore reduce the expansion in credit to the private sector’ Central Bank of Kenya (2011d). 26 Bank of Uganda (2011). 27 International Monetary Fund (2011b). 28 National Bank of Rwanda (2011a). 29 The average maturity of loans in the cleaned loan-level Uganda dataset of Abuka et al. (2015) is 1.5 years. Moreover, it is a characteristic of fully credible regimes that very long rates do not move much with short rates, because inflation expectations are well anchored (Gurkaynak et al., 2007). See also Bulir and Vlcek (2015). 30 While not necessarily relevant in this episode of dramatic tightening, this point is closely related to the argument, e.g., in Woodford (2001), that the substantial smoothing seen in advanced-country monetary policy reaction functions implies a large pass-through to lending rates. A lesser degree of smoothing thus implies lower pass-through. 31 They also identify a bank-balance-sheet channel in which balance sheet conditions of banks influence these effects, with for example poorly capitalised banks transmitting the interest rate changes more strongly. 32 Tanzania’s also relatively closed capital account might have been expected to mitigate the effects of monetary policy shocks on the exchange rate. In practice, the degree to which the nominal exchange rate is allowed to respond to interest rate differentials depended on the set of relatively slowly-moving policies on capital account openness, with a more closed account implying less reaction, and also on policies such as sterilised intervention and more temporary administrative measures that varied substantially through time and which are hard to measure directly. 33 In Uganda, the output gap increased from Q32010 through Q32011 despite positive real interest rates for the first half of that period. Fiscal policy was expansionary during the first part of this period, with a fiscal impulse (the change in the primary balance adjusted for the cyclical position, estimated at 1% of GDP for the Q3:2010-Q2:2011 period) (International Monetary Fund, 2011a). 34 Uganda stabilised somewhat earlier, in August, in this case coinciding more closely with the levelling off of the VIX but also with the earlier beginning of monetary policy tightening phase, which also peaked in October. 35 REER is downloaded from IMF database. The median and percentiles are calculated from REER of Brazil, Chile, Colombia, Czech Republic, Estonia, Ghana, Hungary, India, Indonesia, Israel, Korea, Mauritius, Mexico, Pakistan, Peru, Philippines, Poland, Romania, Serbia, Slovenia, South Africa, Sri Lanka, Thailand and Turkey. 36 OLS results without pre-filtering (available on request) tell a similar story, with if anything a smaller role for the exogenous variables. 37 That is, we create a ‘counterfactual’ based on the predictions of the estimated model given the exogenous variables, rather than using actual lagged values of the endogenous variable. 38 The negative coefficient on the US Rate in the exchange rate regression may be considered puzzling. A higher US rate should lead to a spot depreciation but a subsequent appreciation would be implied by uncovered interest parity. Because this is a regression using monthly data, either sign could be observed. 39 Mishra et al. (2014) find evidence to this effect in a large sample of developing countries. 40 Saxegaard (2006). 41 This evidence suggests that these ‘excess’ reserves may be an equilibrium phenomenon reflecting factors such as risk aversion on the part of banks and structural deficiencies in the functioning of the interbank market. Addressing these deficiencies that create excess liquidity is likely important for the promotion of financial development but, judging from this episode, it does not appear necessary for the transmission of monetary policy. References Abuka C. , Alinda R. , Minoiu C. , Peydro J. , Presbitero A. , 2015 , ‘Monetary Policy in a Developing Country: Loan Applications and Real Effects,’ Working Paper 15/270, International Monetary Fund, Washington, D.C. Adam C. , Maturu B. , Ndung’u N. , O’Connell S. ( 2010 ) ‘Building a Monetary Regime for the 21st Century’, in Adam C. , Collier P. , Ndung’u N. (eds) , Kenya: Policies for Prosperity . Oxford : Oxford University Press . Andrle M. A. , Berg A. , Berkes E. , Morales R. , Portillo R. , Vavra D. , Vleck J. , Money Targeting in a Modern Forecasting and Policy Analysis System: An Application to Kenya , Working Paper 13/239, Washington : International 2013 , Monetary Fund . Andrle M. A. , Berg A. , Morales R. , PortilloWo R. , Vleck J. , 2013 , Forecasting and Monetary Policy Analysis in Low Income Countries (1): Food and Non-Food Inflation in Kenya. Working Paper, International Monetary Fund, Washington, D.C. Bank of Uganda , 2011 , Monetary Policy Statement for October 2011. Berg A. , Charry L. , Portillo R. , Vlcek J. , 2013 , The Monetary Transmission Mechanism in the Tropics: A Narrative Approach. Working Paper, International Monetary Fund, Washington, D.C. Berg A. , O’Connell S. , Pattillo C. , Portillo R. , Unsal F. , 2015 , Monetary Policy Issues in Sub-Saharan Africa, The Oxford Handbook of Africa and Economics: Volume 2: Policies and Practices. Bulir A. , Vlcek J. , 2015 , ‘Monetary Transmission: Are Emerging Market and Low Income Countries Different,’ Working Paper 15/239, Washington: International Monetary Fund. Central Bank of Kenya , 2011 a, MPC Press Release. March. Central Bank of Kenya , 2011 b, MPC Press Release. July. Central Bank of Kenya , 2011 c, MPC Press Release. September. Central Bank of Kenya , 2011 d, MPC Press Release. October. Chinn M. , Ito H. ( 2008 ) ‘ A New Measure of Financial Openness ’, Journal of Comparative Policy Analysis , 10 ( 3 ): 309 – 22 . Google Scholar Crossref Search ADS Christiano L. J. , Eichenbaum M. , Evans C. L. ( 1999 ) ‘Monetary Policy Shocks: What Have We Learned and to What End?’, in Chapter 2 of Handbook of Macroeconomics , Vol 1 . London : Elsevier , pp. 65 – 148 . Part A. Cottarelli C. , Kourelis A. ( 1994 ) ‘ Financial Structure, Bank Lending Rates, and the Transmission Mechanis of Monetary Policy ’, Staff Papers—International Monetary Fund , 4 ( 4 ): 587 – 623 . Google Scholar Crossref Search ADS Davoodi H. , Dixit S. , Pinter G. , 2013 , Monetary Transmission Mechanism in the East African Community: An Empirical Investigation. Working Paper, International Monetary Fund, Washington, D.C. Dotsey M. , Reid M. ( 1992 ) ‘ Oil Shocks, Monetary Policy, and Economic Activity ’, Federal Reserve Bank of Richmond Economic Review , 78 : 14 – 27 . Friedman M. , Schwartz A. ( 1963 ) A Monetary History of the United States . Princeton : Princeton University Press , p. 1963 . Gurkaynak R. , Levin A. , Mardar A. , Swanson E. ( 2007 ) Inflation Targeting and the Anchoring of Inflation Expectations in the Western Hemisphere , Economic Review, San Francisco : Federal Reserve Bank of San Francisco . International Monetary Fund ( 2008 ) Regional Economic Outlook . Washington, DC : Sub-Saharan Africa . April. International Monetary Fund , 2011 a, ‘Uganda—Third Review Under the Policy Support Instrument, Request for Waiver of Nonobservance of an Assessment Criterion, and Request for Modification of Assessment Criteria.’ Washington, D.C. December 21. International Monetary Fund , 2011 b, ‘Tanzania: Letter of Intent, Memorandum of Economic and Financial Policies, and Technical Memorandum of Understanding,’ December 23. International Monetary Fund , 2012 , Annual Report on Exchange Arrangements and Exchange Restrictions. Washington, D.C. Retrieved 2011, from http://www.imfareaer.org/Areaer/Pages/Home.aspx International Monetary Fund , 2015 , Evolving Monetary Policy Frameworks in Low-Income and Other Developing Countries – Staff Report, Washington, D.C. October 23. Kasekende L. , Brownbridge M. ( 2011 ) ‘ Post-Crisis Monetary Policy Frameworks in Sub-Saharan Africa ’, African Development Review , 23 ( 2 ): 190 – 201 . Google Scholar Crossref Search ADS Li B. , O’Connell S. , Adam C. , Berg A. , Montiel P. , 2016 , VAR Meets DSGE: Uncovering the Monetary Transmission Mechanism in Low-Income Countries, Working Paper 16/90, International Monetary Fund, Washington D.C. Mishra P. , Montiel P. , 2013 , ‘How Effective is Monetary Transmission in Low Income Countries? A Survey of the Empirical Evidence’ , Economic Systems , 37 ( 2 ): 187 – 216 . Google Scholar Crossref Search ADS Mishra P. , Montiel P. , Pedroni P. , Spilimbergo A. ( 2014 ) ‘ Monetary Policy and Bank Lending Rates in Low-Income Countries: Heterogeneous Panel Estimates ’, Journal of Development Economics , 111 : 117 – 31 . Google Scholar Crossref Search ADS Mishra P. , Montiel P. , Spilimbergo A. ( 2012 ) ‘ Monetary Transmission in Low-Income Countries: Effectiveness and Policy Implications ’, IMF Economic Review , 60 : 270 – 302 . Google Scholar Crossref Search ADS Montiel P. , Adam C. , Mbowe W. , O’Connell S. , 2012 , Financial Architecture and the Monetary Transmission Mechanism in Tanzania. Working Paper, International Growth Center. National Bank of Rwanda , 2011 a, The BNR Key Repo Rate Increased From 6% to 6.5%. O’Connell S. , Pattillo C. , Portillo R. , Unsal D. , 2015 , Signaling and the Transmission of Monetary Policy: The Rold of the Policy Regime, Unpublished Manuscript (Washington: International Monetary Fund). Romer C. , Romer D. ( 1989 ) ‘Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Friedman’, in NBER Macroecononomics Annual , vol. 4 . Cambridge : National Bureau of Economic Research , pp. 121 – 84 . Saxegaard M. , 2006 , Excess Liquidity and Efectiveness of Monetary Policy: Evidence from Sub-Saharan Africa. Working Paper, International Monetary Fund. Shapiro M. D. ( 1994 ) ‘Federal Reserve Policy: Cause and Effect’, in Mankiw N. G. (ed.) , Monetary Policy . Chicago : University of Chicago and NBER . Stone M. , 2003 , “‘Inflation Targeting Lite”’, WP/03/12, Washington: International Monetary Fund Summers L. ( 1991 ) ‘ The Scientific Illusion in Empirical Macroeconomics ’, The Scandinavian Journal of Economics , 93 ( 2 ): 129 – 148 . Google Scholar Crossref Search ADS Woodford M. , 2001 , “‘Monetary Policy in the Information Economy”’ in Economic Policy for the Information Economy: Proceedings of the Federal Reserve Bank of Kansas City Annual Symposium Conference Jackson Hole, Wyoming, August 30–September 1: 297-370. © The Author(s) 2018. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Journal of African EconomiesOxford University Press

Published: Jun 1, 2019

There are no references for this article.