7 Introduction to Part II

Andrew Berg, and Rafael Portillo
Published Date:
April 2018
  • ShareShare
Show Summary Details
Andrew Berg and Rafael Portillo 

The chapters in Part II reflect our efforts over the last few years to use simple dynamic general equilibrium models to analyse monetary policy issues facing sub-Saharan African (SSA) countries. A few words on the genesis of this research agenda will help clarify, we hope, both the approach and the choice of topics. Our journey started when we were at the ‘policy wing’ of the IMF’s African department and then continued at the development macroeconomics division of the research department.1 These divisions can be thought of as providing support to IMF country teams, and indeed many teams would approach us with thorny policy issues they were facing, either in the context of fund surveillance, or programme design and review. When faced with specific requests, our task would be to offer general policy recommendations, drawing on the policy and academic literature.

Policy thinking typically centred around four broad areas: (i) how to analyse policy under existing monetary frameworks in SSA, (ii) how to think about supply shocks and the adequate policy response, (iii) how to think about exchange rate management via interventions in the foreign exchange (FX) market, in countries with some degree of exchange rate flexibility, and (iv) how to support the authorities in their move to modernize their policy frameworks and what kind of analytical approaches are best suited to these countries. The chapters in Part II mainly cover issues (i)-(iii), while issue (iv) is discussed extensively in Part III.2

As we thought about these policy issues, we concluded that the analysis would benefit enormously from greater use of dynamic stochastic general equilibrium (DSGE) models, especially of the New Keynesian variety, though suitably adjusted to incorporate structural features of low-income countries (LICs).3 These sorts of models dominate monetary policy analysis in the advanced-country context, despite having important detractors. In the LIC policy world their use has been much scarcer, and this scarcity puts their deficiencies in perspective. Indeed, we found and continue to find other approaches wanting. Financial programming, for example, which is the traditional approach used in IMF policy work, emphasizes linkages across sectors in the economy, including monetary, fiscal, and balance of payment accounts.4 In order to perform quantitative policy analysis, however, the framework is complemented with simple behavioural relations, which tend to be primarily focused on monetary stocks and flows and are not suited for meaningful discussions about monetary policy.

Yet another approach is to use simple ad-hoc models, of the type presented in Agenor and Montiel (2008). These can be useful, but in our opinion they often do not sufficiently emphasize the dynamic nature of macro-fluctuations and the importance of expectations in the analysis of monetary policy. Moreover, lack of micro-foundations makes it more rather than less difficult, as is often assumed, to clarify how LICs differ from emerging markets and advanced countries. It also makes it difficult to place the LICs policy debate within the context of the burgeoning academic and policy literature on monetary policy, which over the last twenty years has relied almost exclusively on the New Keynesian framework. Although the latter—at least in its canonical form—misses some of the relevant mechanisms for thinking about policy in LICs, it provides many useful insights for thinking about monetary policy. Our approach has therefore been to build on the latter approach, and to extend or revise these models as needed.

A related observation is that there is a continuity of themes, and in some cases somewhat similar model structures, between Part II and Part III. What distinguishes the two is that the focus in Part II is mainly on qualitative insights rather than quantitative results, and if there are quantitative results, as in Chapter 11, they tend to focus on a narrow dimension of the data. The chapters here are thus meant to provide general guidance, rather than the country-specific policy prescriptions found in Part III. This distinction is also reflected in the approach to modelling. The models in Part II are fully micro-founded, and they tend to be stylized. This is because they are meant to help illustrate how introducing a particular LIC feature into an otherwise standard model may (or may not) influence the policy prescriptions or the effects of policy typically found in the literature. In our view, this type of comparative analysis is most useful if carried out in micro-founded models, as the effort of deriving mechanisms from first principles helps sharpen the discussion.

As previously mentioned, the chapters in Part II introduce a variety of LIC-specific features into otherwise standard New Keynesian models. Some are relatively well known, such as limited influence by LICs over their terms of trade, limited access to international capital markets, and exposure to volatile sources of external funding. Some are related to the broad policy framework, for example the central role of FX interventions, or the issue of coordination with fiscal policy. Other features have not featured prominently in macro models used in LICs, at least until recently. These include pervasive information incompleteness faced by both the public and the private sectors, proximity to subsistence and its implications for the supply side of the economy, limited participation in domestic financial markets and its implications for the effects of various domestic and external shocks, and portfolio balance effects as the channel through which sterilized FX interventions affect the real economy. A summary of the policy issue and the relevant modelling assumptions in each chapter is provided below.

1 Incomplete Information and the Analysis of Existing Monetary Policy Arrangements in SSA

As discussed extensively in Chapter 1, many SSA countries with some degree of exchange rate flexibility have money targeting frameworks, which are also reflected in the design of IMF-supported programmes in the region, i.e., in the conditionality related to monetary policy. In practice there is considerable flexibility, and money target misses are common, partly because central banks wish to avoid either excessive interest rate volatility or large misalignments in the level of short-term rates. Such flexibility, however, complicates the analysis of monetary policy, including under fund programmes. As has become clear from our many interactions with IMF country teams over the years, the latter often struggle to reconcile the broader discussion of the adequate monetary policy stance with the programme conditionality. The two chapters summarized below reflect an attempt to provide guidance on this issue.

Chapter 8 reflects relatively early efforts on our part to engage constructively with prevalent practices by deriving the conditions under which paying some attention to money targets (with flexibility vis-à-vis target misses) could make sense from a policy perspective. In the context of a simple New Keynesian model, we show that some adherence to money targets can emerge endogenously when the central bank wishes to implement a Taylor-type rule, a standard monetary policy prescription, but cannot perfectly observe the state of the economy or the relevant interest rate for private sector decisions. The latter is a short cut for the idea that financial markets are insufficiently developed to be able to easily infer the relevant risk-free interest rate from market outcomes. The degree of money target adherence, which we denote λ, emerges endogenously as part of a signal extraction problem faced by the central bank. We derive an analytical solution for the optimal degree of target adherence, which we denote λ*, and show that it depends on structural parameters of the model, including the deviation (noisiness) between observed and relevant interest rates, and the volatility of indicators of the state of the economy and various real shocks. Our approach also yields a simple rule for the design of money targets: these should be thought of as optimal forecasts of future money demand, consistent with the desired stance of policy, i.e., the interest rate level implied by the Taylor rule.

The previous discussion could be interpreted as implying that the current state of affairs with regards to monetary policy in SSA is optimal. As we continued to engage with countries attempting to use these sorts of frameworks, we came increasingly to emphasize that their flexibility reflects some costly disarray in the implementation of monetary policy, with lack of clarity about what the operational target is or should be, and lack of instruments or frameworks for controlling short-term financial conditions. Chapter 9 looks at such ‘policy implementation errors,’ and analyses their implications for the effectiveness of policy in the same model, except this time it is the private sector that faces information incompleteness. Agents in the economy have to infer the intended stance of policy solely from money market developments, a reasonable assumption given the typical absence of clear communication by central banks in LICs. In such an environment we show that the effectiveness of policy decisions can be greatly reduced when implementation errors are pervasive. As a result, it may appear as though monetary policy has weak effects on output and inflation, and indeed this is what many empirical studies seem to find, in the case of Africa. Chapters 5 and 6 argued that these results could reflect limitations in the use of empirical tools for assessing the monetary transmission mechanism. Here it is shown that they could also reflect policy shortcomings, which are easier to fix than the deeper structural issues that are often put forward in the literature.5 A policy corollary from this analysis is that de facto flexibility with regards to money targets is no substitute for clear and transparent frameworks for policy design and policy implementation, with the latter centred around the explicit control of short-term interest rates.

From an analytical perspective, the lesson we draw from Chapters 8 and 9 is that the explicit modelling of information incompleteness can be useful for studying monetary policy issues in LICs, especially given the data shortcomings and the overall policy opacity.

2 Supply-Side Shocks and the Structure of the Economy

Understanding how supply shocks propagate through the economy, and whether monetary policy should respond to the resulting inflationary pressures, is one of the fundamental questions in monetary policy. Not surprisingly, interest in the topic tends to spike when there are large supply-side disturbances, for example as during the oil shocks of the 1970s, and more recently during the 2007–08 international food and fuel commodity crisis. As economists working on Africa during the latter period, we were thus confronted with this difficult policy question.

Standard policy advice, including at the IMF, calls for central banks to respond only to ‘second-round’ effects—spillovers from food prices to wages and core inflation—and to accommodate first-round effects. The latter term is rather vague, but is meant to describe the inherent inflationary pressure from the shock, which stems from the need for relative prices to adjust. Relative to other regions, SSA stands out, however, by the perceived prevalence and magnitude of supply shocks, and in the case of shocks to food prices, the effect is amplified by the large weight of food in the CPI. Does the standard advice still apply under these conditions?

Chapter 10 attempts to provide a formal definition of first-round effects, in the context of a stylized New Keynesian open economy model, and examines whether international food price shocks are more inflationary in countries with a larger food share in consumption, denoted αF. This issue was of great concern in Africa during 2007–08, as large increases in international food prices coincided with a marked increase in inflation in many countries in the region.

The analysis in the chapter reveals that much depends on the country’s access to international capital markets. If a country has ample access, then ap is indeed the key parameter determining the magnitude of first-round effects. However, if a country has limited access to capital markets, then the key parameter is the country’s trade deficit in agricultural products (αFκF in the model), rather than αF. The reason is that under limited international capital mobility, it is the income effect from the shock that determines the magnitude of the relative price adjustment, and this effect is proportional to the trade balance.6 As αFκF is on average small in SSA countries, and it can be argued that access to international capital markets is rather limited, the chapter makes the provocative claim that inflationary pressures from international food price shocks are small in SSA, at least on average.

Although the stylized nature of the model cautions against taking this result too literally, the analysis suggests that other factors may have accounted for the inflationary pressures observed back in 2008, most notably the stance of monetary and fiscal policy. This issue is taken up again in Chapter 15 in the case of Kenya. The results in the chapter also hint at the complexities involved in formally deriving ‘first-round’ effects, as it shows that they depend on equilibrium changes in nominal exchange rates, which themselves depend on the specific monetary policy response that keeps core inflation stabilized. Such complexity limits the applicability of the ‘first-round’ concept for policy analysis. The vagueness mentioned above is, to some extent, inevitable.

Chapter 11 studies how the structure of African economies affects the proper ties of inflation and the monetary policy response. It does so by introducing subsistence requirements in food consumption in a simple New Keynesian model with two sectors, food and non-food, and with shocks to food productivity (supply shocks) and to aggregate demand. Subsistence requirements are used in growth models to generate structural transformation: they imply that, as overall productivity in the economy increases, the share of food in consumption (and the share of factors of production allocated to agriculture) decreases endogenously. Here, instead, subsistence requirements are used to analyse how the properties of inflation change, also endogenously, across levels of income. The chapter shows that greater proximity to subsistence (lower levels of development) makes inflation predominantly supply-driven and more volatile, for purely structural reasons, as the share of food in consumption is larger and there is more limited economy-wide ability to adjust to supply shocks. In spite of this greater exposure to supply shocks, the standard policy prescription remains: stabilizing core (non-food) inflation is optimal from a welfare perspective. Moreover, the model analysis reveals that stabilizing overall inflation, instead of core, is costlier in poor countries, as monetary policy needs to generate large output volatility to stabilize an inherently volatile inflation.

3 The Macroeconomic Implications of Reserve Accumulation and FX Intervention Policies

The discussion so far implicitly assumes that monetary policy is about the choice of one policy instrument, the short-term interest rate (or the monetary base). In SSA countries, and developing countries more generally, things are more complex. A key missing element, in countries with some degree of exchange rate management, is FX interventions, both to achieve a desired level of reserves, and to limit exchange rate volatility or protect competitiveness. Until recently there was no standard framework for thinking about interventions and whether, and if so how, they should be coordinated with interest rate policy or the broader macro policy stance.7 This is reflected in the sometimes erratic policy responses of central banks, as well as in the analysis of IMF country teams, who often struggle when assessing what the effect of those interventions may be, and whether interventions are desirable or compatible with the broader macro framework. The chapters below develop models that fill this gap.

Chapter 12 focuses on FX interventions in the context of surges in aid flows, an important issue in many SSA countries, especially in the early 2000s. Governments receiving aid disbursements would initially deposit the FX at the central bank. As they drew down on these deposits to spend on local goods and services, liquidity would be created. Against this backdrop, central banks felt they were confronted with a policy trilemma: if they sold the aid-related FX to sterilize the liquidity, the real exchange rate would appreciate; if instead they sold government bonds or issued their own sterilization instruments, interest rates would increase; and if they did not sterilize, inflation would rise. Looking at various countries in SSA, Berg et al. (2007) found that many central banks ended up keeping the aid proceeds as reserves, with the experience of Uganda (which is also summarized in the chapter) providing a case in point. It was the challenge of thinking about this policy choice that led to the analysis in the chapter.

Using a small open economy model that includes a specification of the government sector and a rule for FX interventions, Chapter 12 shows that the policy trilemma is more than just an issue of liquidity management by the central bank. The underlying issue is that the real effects of the aid are very different depending on whether the FX proceeds are accumulated as reserves or not. Reserve accumulation implies that the public sector is using the aid twice, once as higher government spending and once as reserves. If access to international capital markets is limited, then the private sector ends up being crowded out, either through higher interest rates (if the reserve accumulation is sterilized) or through the inflation tax (if it is not). The core policy issue is the need for coordination between the response of the fiscal authorities and those of the central bank: either the aid is used, in which case higher government spending should be associated with higher FX proceeds being made available to the economy, or it is not used, in which case higher public savings can help support higher FX reserves.

What the above analysis makes clear is that FX intervention policies have real effects that are different from the ‘traditional’ effects of monetary policy, stemming from the control of short-term nominal variables. The last chapter in Part II explores this issue in more detail. It extends a standard open economy New Keynesian model to include a rule for sterilized FX interventions operating alongside interest rate policy, and portfolio effects of FX intervention policies. The portfolio effects stem from the wedges or premia on the expected returns of different assets and liabilities which depend on the relative supply of the financial assets themselves. By affecting the relative holdings of central bank paper in the financial system, sterilized interventions influence nominal and real exchanges and therefore can also influence output and inflation.

The existence of multiple instruments (interventions and interest rates) and channels of transmission increases the range of policies, for example the possible co-existence of an interest-rate-based inflation targeting regime with a managed float. Chapter 13 shows that there can be benefits to these hybrid regimes, for example when the economy is hit with temporary shocks to foreign interest rates or risk premia: relying on FX interventions helps insulate domestic interest rates, and the economy, from these external shocks. In other words, the use of FX interventions can increase monetary policy autonomy with regards to external developments. Not all FX rules are equal, however. ‘Leaning against the wind’ rules, which attempt to reduce the pace of appreciation/depreciation, without targeting a specific level for the exchange rate, are preferable to rules that do target an exchange rate level, as the former are more robust to weak portfolio balance effects and are less likely to result in the depletion of reserves. More generally, the latter raise the risk of pursuing multiple, inconsistent objectives, and can lead to runs on the currency and macro instability.

This concludes the summary of the chapters in Part II. We hope it has shown that the analysis of monetary policy issues in LICs, and SSA in particular, can benefit tremendously from building dynamic stochastic general equilibrium models that capture important features of these economies, including complex policy environments, and by using them in a sensible manner, i.e., not mechanically but rather to build intuition about the channels involved. Many issues remain to be analysed, for example on the use of macro-prudential tools in LICs, or the distributional effects of macro policies, just to name two broad policy areas.8 Studying these issues in meaningful ways will also require further development of these models along the relevant dimensions, for example by carefully modelling the financial sector, as done to some extent in Chapter 17, and by introducing greater heterogeneity among households and firms. This is left for future work.


    Agenor, P. andMontiel, P. (2008). Development Macroeconomics. Princeton, NJ: Princeton University Press.

    Berg, A.,Yang, S. C., andZanna, L. F. (2014). Modeling African Economies: A DSGE Approach. In C.Monga andJ. Y.Lin (Eds.), Oxford Handbook of Africa and Economics: Context and Concepts. Oxford: Oxford University Press.

    • Search Google Scholar
    • Export Citation

    Berg, A.,AiyarS.,HussainM.,RoacheS.,MirzoevT., andMahone, A. (2007). The Macroeconomics of Scaling Up Aid: Lessons from Recent Experience. IMF Occasional Paper 253. Washington, DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

    Clarida, R.,Gali, J., andGertler, M. (1999). The Science of Monetary Policy: A New Keynesian Perspective. Journal of Economic Literature, 37 (4), 1661–1707.

    • Search Google Scholar
    • Export Citation

    Gali, J. andMonacelli, T. (2005). Monetary Policy and Exchange Rate Volatility in a Small Open Economy Model. Review of Economic Studies, 72, 707–34.

    • Search Google Scholar
    • Export Citation

    Mishra, P.,Montiel, P., andSpilimbergo, A. (2012). Monetary transmission in low-income countries: effectiveness and policy implications. IMF Economic Review, 60, 270–302.

    • Search Google Scholar
    • Export Citation

    Polak, J. J. (2005). The IMF monetary model at forty. In J.Boughton (ed.), Selected Essays of Jacques J. Polak, 1994–2004 (pp. 209–26). New York and London: Sharpe.

    • Search Google Scholar
    • Export Citation

    Unsal, F. andRubio, M. (2017). Macroprudential Policy, Incomplete Information and Inequality: The case of Low-Income and Developing Countries. IMF Working Paper 17/59. Washington, DC: International Monetary Fund.

    • Search Google Scholar
    • Export Citation

In one case (Andrew Berg), the research agenda goes back a bit further, mainly his time at the strategy and policy review department of the IMF. This wandering through several departments is typical in the career of IMF economists, as the institution tends to value ‘fungibility and general expertise.


Issue (iv) also reflects another function provided by these divisions, especially the development macroeconomics division of the research department, which is to directly engage with various central banks in SSA on these issues through technical assistance and capacity development projects.


For a discussion of the standard closed and open economy New Keynesian Models, see Clarida, Gali, and Gertler (1999) and Gali and Monacelli (2005), respectively. On their application to low- income countries, see Berg et al. (2014).


See Polak (2005), among others.


Under ample access to international capital markets, it is the substitution effect that drives the magnitude of the relative price change. Substitution effects are proportional to αF.


This is much broader than the issue of whether interventions should be sterilized or not.


A recent contribution on the model-based analysis of macro-prudential policy in low-income countries is Unsal and Rubio (2017).

    Other Resources Citing This Publication