Chapter

II Analytical Tools for Policy Formulation and Program Design

Author(s):
Charalambos Christofides, Atish Ghosh, Uma Ramakrishnan, Alun Thomas, Laura Papi, Juan Zalduendo, and Jun Kim
Published Date:
September 2005
Share
  • ShareShare
Show Summary Details

An IMF-supported program is a package of policy measures which, combined with approved financing, is intended to achieve certain economic objectives.3 In essence, therefore, a program is defined by its objectives, the link between those objectives and policy instruments, and thus the specification of macroeconomic and structural policies. This section considers the process and analytical tools used for establishing the link between policies and objectives in the formulation of IMF-supported programs.

One approach to policy formulation would be for national authorities and IMF country teams to develop a comprehensive macroeconomic model linking policies to targets. This model could then be inverted to derive the policies necessary to achieve them. If the IMF was confident that the implied policies would be implemented, it would support a program that predicts that sufficiently ambitious targets would be achieved.

While such an approach would have a number of advantages—ensuring consistency, illustrating the effects of alternative policy mixes, and identifying intertemporal trade-offs between financing and adjustment—empirical and practical considerations make the use of comprehensive macroeconomic models implausible in most cases.4 Instead, therefore, national authorities and IMF country teams typically rely on a variety of approaches to help formulate macroeconomic and structural policies. For the purposes of discussion, it is useful to consider the process of policy formulation for short-run objectives (such as macroeconomic stabilization and external adjustment) separately from the longer-term goals of ensuring debt sustainability, reducing vulnerabilities, and raising the growth potential of the economy—though, of course, these are dynamically linked.

Macroeconomic Stabilization and External Adjustment

In formulating their economic program, national authorities have a number of different instruments—the exchange rate regime, monetary policy, fiscal policy, and structural measures. While such policy prescriptions would be consistent with most open economy macroeconomic models, the specific policy content of the authorities’ program naturally depends upon the country’s characteristics and the circumstances it is facing. Thus, if Keynesian effects are likely to be important, then the effect of fiscal consolidation on activity and output growth would need to be taken into account. Likewise, the pace at which disinflation should be targeted—and the choice of nominal anchor—should be viewed against the benefits for growth of macroeconomic stability, the possible need to adjust administered prices in the economy, and realistic expectations regarding fiscal policy.5 Since no single model is universally applicable, national authorities—and IMF country teams in advising them—must draw on an array of small econometric models and single equation estimates (including existing analytical work undertaken by research departments in central banks, ministries of finance, and private think tanks), as well as economic judgment for formulating macroeconomic and structural policies.

Box 3.1.The Anatomy of Program Design: Indonesia, 2000

Policy formulation for Indonesia’s 2000 EFF arrangement provides a typical example of the process of program design. The preparation of a macroeconomic framework started with preliminary output and price projections, followed by projections for the fiscal, external, and monetary sectors. Given the linkages among the various sectors, achieving internally consistent and economically meaningful projections required an iterative rather than a recursive process. The steps involved are summarized below.

Real sector. Projections were expenditure based, with the real GDP growth rate and consumer price inflation assumptions reflecting program objectives. Public consumption and investment were obtained from the fiscal accounts, while private consumption and investment were based on the expected recovery of the banking and corporate sectors. Export and import volume growth rates were obtained from the external accounts, while the change in inventories was derived as a residual.

Fiscal sector. The targeted overall balance (a performance criterion under the IMF-supported program) sought to balance the twin objectives of supporting recovery and reducing the public debt, while being mindful of the available financing (to limit base money growth consistent with the inflation target, the program did not allow for domestic bank financing; the program also established limits on arrears accumulation). On the revenue side, oil and gas revenue projections were derived using the IMF’s World Economic Outlook oil price assumptions and the assumed program exchange rate. Non-oil revenues were derived from the projected nominal GDP growth with adjustments for policy implementation (such as better revenue collection and higher tax ratios). On the expenditure side, the projections were made using a combination of nominal GDP growth and historical expenditure ratios with adjustments for policy implementation (such as lower payments on subsidies). The projections were also influenced by the upcoming need for implementing fiscal decentralization.

External sector. The components of the external current account were projected based on the World Economic Outlook projections for oil prices, import deflators, and trading partners’ growth rates; program exchange rate assumptions; and estimates of price and income elasticities for exports and imports. The capital account was derived based on estimates of official capital flows from various multilateral and bilateral sources, estimates of private capital flows (including the projected returns from corporate and bank restructurings), and exceptional financing items. The net international reserves (NIR, performance criterion) accumulation target was set to zero for the first year of the program. A small recovery in NIR was targeted for subsequent years.

Monetary sector. Attempts at estimating traditional money demand functions to arrive at a path for the monetary variables did not yield stable results. Therefore, the monetary projections were based on assumed monetary ratios and historical benchmarks. Among the components of Bank Indonesia’s balance sheet, base money (an indicative target) was derived from projections of currency in circulation and deposits (bank and nonbank). Currency in circulation was derived by multiplying the rupiah broad money by the long-term trend of the ratio of currency to rupiah broad money. Bank deposits were derived by applying the reserve requirement ratios to rupiah M2, and non-bank deposits were held constant. On the assets side, consistent with the balance of payments projections, NIR for the initial program year was assumed to be constant so as not to exert an expansionary influence on base money; in the outer years, accumulation was allowed. Net domestic assets (performance criterion) was derived residually from base money and NIR. In the monetary survey, rupiah broad money was derived by applying its trend growth rates, and private credit was assumed to be in line with the nominal GDP growth rate.

The program thus developed is essentially defined by a core set of macroeconomic projections on real GDP growth, inflation, the current account, and the balance of payments. In turn, these variables both influence, and are influenced by, monetary, exchange rate, and fiscal policy instruments. Thus inflation and growth will be important inputs into fiscal revenue and expenditure projections, but the size of the deficit may have a bearing on economic activity, and its financing on inflation and interest rates. Likewise, the monetary policy stance has implications for prices and output growth, but real growth, in turn, is likely to affect the demand for money.

The mutual dependence of instruments and targets means that the modeling process is usually iterative and often quite complex (Box 3.1). A key concern is ensuring consistency of the macroeconomic framework and coherence of the policy stance across instruments to meet program objectives. Financial programming is used as a general approach6 to inform and tie together the various sectors in a consistent manner, while incorporating country-specific factors.7 In this fashion, not only does financial programming serve as an ex ante consistency check on important financial aggregates, it also provides an ex post monitoring tool.8

A key characteristic of this approach is that it allows policies to be adjusted and reformulated in a dynamic manner in light of outcomes.9 Policy formulation thus extends well beyond the Executive Board approval of an IMF arrangement. Indeed, program reviews are intended to offer the opportunity for country authorities and IMF staff to reassess their initial assumptions and the progress achieved during the first few months of the program, including reasons why program objectives may be deviating from targets, with the forward-looking aspect of IMF reviews allowing for policy adjustments to help ensure that program objectives are achieved.10

The need for frequent reassessments of policies in light of outcomes is especially acute in capital account crises. Although these programs are, at one level, little different from more traditional programs—typically targeting some external adjustment (on average, about 1.1 percent of GDP; Table 3.1)—their salient feature is the large and sudden capital outflows that force much larger-than-envisaged adjustments of the current account balance (on average, 8.5 percent of GDP), with pervasive macroeconomic consequences.11 In particular, the timing and magnitude of the capital outflows are very difficult to predict—indeed, existing models of capital flows perform very poorly even in noncrisis situations.12 These flows and the attendant exchange rate movements interact with domestic balance sheet exposures, potentially altering the magnitude, and even the sign, of the effects of economic policies.13

Table 3.1.Projected and Actual Current Account Adjustment in Capital Account Crisis Programs(In percent of GDP)
Approval YearCrisis YearCurrent Account Adjustment
ProjectedActual
Argentina200020020.112.0
Brazil199819990.6-0.6
Indonesia199719980.56.0
Korea199719982.514.4
Mexico199519953.76.5
Russia199619990.012.1
Thailand199719982.014.8
Turkey199920010.37.2
Uruguay200020020.24.4
Average1.18.5
Sources: IMF, MONA and WEO databases; and IMF staff estimates.
Sources: IMF, MONA and WEO databases; and IMF staff estimates.

Partly in response, the IMF has been developing a balance sheet approach to understand the mechanisms underlying these stock shifts.14 From the perspective of this approach, a financial crisis occurs when a plunge in demand for financial liabilities takes place in one or more of the sectors—creditors may lose confidence in the sovereign’s ability to service its debt, in the banking system’s ability to meet deposit outflows, or in corporations’ ability to repay bank loans and other debt—ultimately spilling on to the balance of payments. Since most emerging market countries borrow in foreign currency, some sectors in the economy have foreign exchange risk. The key insight is that the maturity structure and distribution of those liabilities across domestic balance sheets, as well as the interrelationships between balances among residents, may have important bearing on the country’s vulnerability to a shift in confidence. The balance sheet analysis can help pinpoint the source of balance of payments disequilibrium and, possibly, the form of intervention that might succeed in containing the crisis.

While a useful addition to the analytical toolkit, it is important to recognize the limitations of the approach. First, although it can help identify vulnerabilities, it cannot predict either the timing or the magnitude of a possible crisis and the capital out-flows.15 Second, though some balance sheet structures may be more resilient than others, as long as the country as a whole has foreign exchange exposure, some balance sheet within the economy faces risks that cannot be diversified away. Finally, there are a number of difficulties in the practical application of the framework, particularly related to the availability of data.

Promoting Economic Efficiency and Output Growth

Enhancing economic efficiency and promoting growth are important goals of IMF-supported programs, particularly among low-income countries.16 Although the economics profession is far from reaching consensus on what drives growth, several conclusions emerge from various studies. First, most studies agree that macroeconomic stabilization is a sine qua non for sustained output growth and for reaping the benefits of any structural reforms, possibly because high and volatile inflation might lessen the value of price signals and distort the allocation of resources.17 Second, while there is less agreement on the best sequence for reforms, a widely accepted view is that stabilization should precede trade18 or financial sector19 liberalization, particularly if these can adversely affect stabilization efforts by reducing trade-related revenues or raising the costs of public sector funding. Moreover, domestic financial markets should be liberalized—with well-supervised prudential regulations in place—before the capital account to ensure an efficient allocation of resources and to limit vulnerabilities. Third, a growing body of literature emphasizes the importance of sound institutions for sustained output growth. At the same time, the variety of judicial systems and institutions in strong performing economies belies the idea that any single approach works best in all countries.20

Beyond these general prescriptions, there are four main analytical tools for understanding the determinants of activity and output growth. First is the demand side assessment—that is, a decomposition of the expected growth into private consumption, investment, government spending, and the current account balance. While this does not provide a model of potential output, it does provide a check on whether the growth projection is consistent with other program parameters, for instance fiscal adjustment. Second, mechanical univariate approaches, such as Hodrick-Prescott filters, may be useful input to medium-term growth projections.21 Third, estimating the aggregate production function may also serve to model growth and discipline projections. However, even though this provides a model for potential output growth, it requires data that are not readily available (e.g., capital stock data), as well as assumptions regarding competition in factor markets or estimates of factor utilization; of course, growth of potential output need not translate into actual output growth if demand is lacking or economic inefficiencies abound. Fourth, growth regressions can be used to map country characteristics (including the availability of factors of production, such as physical and human capital, and structural characteristics, economic policies, and institutions) into its expected growth performance based on cross-country experience. Such regressions perform best over medium-term horizons—usually a five-year period—where business cycle movements and the effects of temporary shocks are averaged out. Their major drawbacks are their data requirements, and the inherent difficulties of quantifying some of the explanatory variables, such as the quality of institutions. One approach would be to examine the association between growth and specific measures in similar (possibly neighboring) countries. These comparisons could usefully include data on medium-term growth rates in countries that are facing similar challenges and are situated at similar stages of development, which in turn would serve to further discipline projections.

A growth model would also allow an assessment of whether the assumed acceleration in growth is realistic and whether the structural reforms embodied in the authorities’ reform program could plausibly lead to such an acceleration. At the same time, it needs to be recognized that it is enormously difficult to map specific measures into the structural indices typically employed in growth regressions. Moreover, while the authorities may draw on cross-country experience to identify broad areas where reforms could bolster growth, they must rely on their own country-specific knowledge to determine the growth bottlenecks that are critical for their country. Even when there is agreement on which reforms might contribute to better economic performance, a further difficulty lies in determining the impact of these reforms on growth—specifically, whether the beneficial effects are likely to peter out quickly or to have a lasting effect on a country’s growth performance. Empirical research suggests that various economic measures—macroeconomic stability, an enabling business environment, trade liberalization, fiscal sustainability, and financial sector development—boost output growth, but in some cases—such as trade liberalization and fiscal sustainability—the long-run effect on the growth rate is weaker (Box 3.2). It would therefore be important to distinguish between immediate and lasting effects on growth rates in assessing the impact of structural reforms on the country’s growth performance.

Despite the availability of these analytical tools, and notwithstanding their shortcomings, a review of a sample of staff reports over a six-year period shows that they typically make limited use of these tools (9 out of 20 staff reports used one or another of the above described techniques, and in almost all cases only once over the six-year period; see Box 3.3). As discussed below, greater use of analytical tools could discipline medium-term growth projections embodied in programs as well as helping to identify some of the impediments to growth pertinent to the particular country.

Box 3.2.Permanent and Temporary Growth Effects from Economic Policies

A cross-country growth equation representing five clusters of economic policies suggests that improving each of these clusters by one standard deviation leads to improvements in growth rates (see Table A) that range from 0.3 percent to 0.6 percent a year (see Zalduendo (2005) for a discussion of the use of factor analysis to capturing different dimensions of economic policy). The five clusters were derived by applying factor analysis to different economic policy indicators. These empirical results use an unbalanced panel of five-year periods between 1981 and 2000. The two clusters of macroeconomic policy are viewed as proxies to economic stabilization and fiscal sustainability. The three clusters of reforms represent trade liberalization policies, financial sector development, and an enabling environment for private sector activity.

Table A.Effects on Growth Rates
CoefficientStandard DeviationAnnual Growth Effect
Business environment0.040.140.51
Financial sector development0.020.160.28
Economic stabilization0.060.070.42
Trade liberalization0.070.070.49
Fiscal sustainability0.070.080.58

Is the growth pay-off from a given improvement in economic policies permanent? The econometric results (see Table B) suggest that growth effects from sound policies are lasting, but that some policies have a more lasting impact than others. This conclusion is derived by comparing three regressions. The first regression includes only contemporaneous measurements of economic policy clusters. The second regression includes only lagged indicators (i.e., the average of the preceding five-year period). The last regression combines contemporaneous indicators and the lagged five-year period for each economic policy regressor. The coefficient estimates in the first equation are positive (better policies support growth) and statistically significant. The conclusions from the second equation are similar, albeit less robust—only lagged macroeconomic stabilization and trade liberalization are statistically important for growth. In contrast, the last equation has positive coefficient estimates on contemporaneous indicators of policy clusters and negative estimates in the lagged indicators. While the sum of the corresponding statistically significant contemporaneous and lagged coefficients is still positive, it is weaker than the contemporaneous effect by itself. More precisely, the combined contemporaneous and lagged coefficients for trade liberalization and fiscal sustainability (equation 3) are smaller than those in the specification that has only contemporaneous regressors (equation 2). In sum, even though the observed growth effects are lasting, in some cases they weaken over time.

Table B.Growth Effects1(Dependent variable: growth rate in GDP per capita)
Equation 1Equation 2Equation 3
Economic policy regressors
Business environment0.0359***0.0717***
(3.40)(4.44)
Financial sector development0.0177***0.0248***
(2.91)(3.47)
Economic stabilization0.0639***0.0920***
(4.17)(4.34)
Trade liberalization0.0652***0.1552***
(3.67)(4.19)
Fiscal sustainability0.0720***0.1108***
(4.33)(6.03)
Business environment, lagged0.0097–0.0442***
(0.99)(–2.96)
Financial sector development, lagged0.0067–0.0050
(1.55)(–0.60)
Economic stabilization, lagged0.0680**–0.0004
(2.54)(–0.01)
Trade liberalization, lagged0.0525***–0.0998***
(5.10)(–2.73)
Fiscal sustainability, lagged0.0125–0.0527***
(0.94)(–3.11)
Wald statistic191.57204.55271.63
Standard error of regression0.02130.02300.0208
R-squared0.430.320.45
Number of observations172
Number of different countries61
Note. *** indicates significance at 1 percent level, and ** indicates significance at 5 percent level; t-statistics are in parentheses.

Regression includes a number of nonpolicy regressors, such as initial income level, terms of trade shocks, and indicators of domestic shocks. Some of these regressors serve to control for differences in initial conditions.

Note. *** indicates significance at 1 percent level, and ** indicates significance at 5 percent level; t-statistics are in parentheses.

Regression includes a number of nonpolicy regressors, such as initial income level, terms of trade shocks, and indicators of domestic shocks. Some of these regressors serve to control for differences in initial conditions.

Medium-Term Debt Sustainability

An important use of medium-term growth projections is to inform debt sustainability assessments. Indeed, going beyond flow balance of payments problems, IMF-supported programs are also intended to reduce vulnerabilities to future crises so that a country should emerge with both its public and external debt dynamics on a sustainable path. To assess debt dynamics, the IMF has developed a standardized debt sustainability template.22 The template lays bare the key assumptions underlying the debt sustainability analysis so that their realism can be assessed against a country’s historical experience. The template also applies stress tests to the baseline projection to examine its resilience to shocks and serves to anchor near-term policy recommendations.

Some of the debt sustainability template’s features and limitations are also worth noting, however.23 First, the template articulates debt dynamics under the baseline and stress scenarios, and thus helps arrive at judgments about the sustainability of a given path of debt, but cannot replace the need for such judgments. Second, the template is intended to take account of the main shocks—such as poor growth performance or real exchange rate depreciations—that could result in an unsustainable increase in debt, but not to model the crisis itself. Thus, while the template tracks gross financing needs, it is not well-suited to modeling how liquidity crises manifest since it focuses only on the country’s aggregate net external debt and capital flows. Third, although the template helps discipline projections, it does not specify a particular model or method that country teams should use in making program projections; ultimately, debt sustainability assessments will only be as good as the macroeconomic projections, including for output growth, underlying it.

Box 3.3.The Treatment of Growth in Staff Reports: Theory and Practice

Even though growth projections are not intended to be forecasts (they are conditioned, inter alia, on policy implementation, and reflect a mix of quantitative analysis and judgment reached during discussions between country authorities and IMF staff), the existence of systematic biases are problematic because they provide a poor basis for choosing the macroeconomic policies and distort the assessment of debt sustainability.

What options are available when preparing growth projections?

The options depend on the length of the projection period and the availability of data. One-time factors, sector-specific issues, and cyclical factors play an important role when preparing short-term growth projections. Unfortunately, many of these factors do not lend themselves easily to formal modeling. A detailed demand-side analysis also provides a useful perspective when preparing growth projections. The range of options broadens for medium-term projections: mechanical univariate approaches (HP-type filters), production function approaches, and growth equations. However, these options also have limitations. Univariate-based assessments have difficulty in distinguishing between trend and one-time factors. Production function approaches are intensive on data that are frequently not available, such as capital stock data, or based on accounting exercises that depend heavily on assumptions regarding factor utilization and production function parameters. Growth equations lack a theoretical foundation. An additional difficulty relates to the quantification of the effects of structural reforms on growth. In fact, it is fair to say that individual reforms might have a limited effect on output. More likely, it is the accumulation of sound economic management and structural reforms that strengthens a country’s growth prospects.

A review of selected IMF reports (20 staff reports for Article IVs and use of IMF resources programs, as well as selected issues papers; see table) covering the period 1995–2000 reveals that:

  • Almost half of the reports utilized at least once during the six-year period an analytical framework for growth projections—HP filters and incremental capital output ratio (ICOR) relationships were the most frequently employed techniques;

  • On slightly over half of the sample analytical work was not feasible or not explicitly described in the reports;

  • Links between reforms and growth are rarely analytical, perhaps reflecting the quantification difficulties mentioned above;

  • Most reports provide a demand-side assessment based on a S-I discussion, though these assessments are not always fully explained;

  • Commodity-based countries provide a supply-side analysis (weather and positive shocks); and

  • Other supply-side assessments refer to sector-specific factors, such as developments in the oil sector.

Use of Analytical Growth Frameworks
Production Function
YearHP FilterGrowth EquationGrowth accountingDerivative of production functionOther
Argentina1996x
Armenia1996x1
Central African Republic1998x1
Congo, Rep. of1996x2
Côte d’Ivoire1998x3
Ghana1999x1
Guinea-Bissau1995x3
Guinea-Bissau2000x1
Guyana1998x4
Jordan1998x
Kenya2000x
Kyrgyz Republic2000xxx
Macedonia, FYR1997x1
Madagascar1996x3
Malawi2000x1
Niger2000x2
Philippines1999xxx
Senegal1998x1
Vietnam1999x
Zambia1995x1
Source: IMF Staff Reports, 1995–2000.

Ad hoc (e.g., increase savings and investment through program reforms).

Underlying population growth and total productivity assumptions.

Based on ICOR assumptions.

Report mentions rise in productivity, though no model is discussed.

Source: IMF Staff Reports, 1995–2000.

Ad hoc (e.g., increase savings and investment through program reforms).

Underlying population growth and total productivity assumptions.

Based on ICOR assumptions.

Report mentions rise in productivity, though no model is discussed.

    Other Resources Citing This Publication