The IMF Research Bulletin summarizes key research from the IMF and includes a listing recent publications in different online series.

Abstract

The IMF Research Bulletin summarizes key research from the IMF and includes a listing recent publications in different online series.

The use of conventional vector auto-regression (VAR)-based methods to identify the monetary transmission mechanism (MTM) in low income countries suggests the MTM may be weaker and less reliable than in advanced and emerging market economies. But are structural VARs identified via short-run restrictions fit for this purpose? Are they capable of detecting a transmission mechanism when one exists, under the structural and research conditions typical of these countries? This Q&A article provides brief answers to seven questions on estimating the monetary transmission mechanism in low-income countries.

Question 1. Why does understanding the monetary transmission mechanism matter in low-income countries (LICs)?

The monetary transmission mechanism (MTM) describes the link between monetary policy instruments under the direct control of the central bank—usually a short-term policy interest rate—and the ultimate economic outcomes it is seeking to influence, typically aggregate demand and inflation.

Transmission occurs along multiple channels: directly through the effect of interest rates on private consumption and investment, and indirectly through exchange rates on import prices and external competitiveness; through the quantity and price of credit from the banking and non-bank financial sectors; through asset prices and wealth effects; and through private sector inflation expectations. The transmission of monetary policy signals via these different channels depends on a range of factors including the depth and development of financial markets, the structure and credibility of the monetary policy regime, and the volatility of the domestic and external economic environment.

A fundamental challenge in the conduct of monetary policy is for central banks to develop robust estimates of the speed, direction, and relative strength of transmission of its policy actions through each channel. This is a particularly difficult task in low-income countries where financial markets are thin, the economy is undergoing rapid structural change and policy regimes are seeking to adjust to this change, and where limitations in the quantity, timeliness, and quality of macroeconomic data are widespread. Structural factors that play a role in ensuring the transmission mechanism in LICs may be relatively weak compared to those in advanced and emerging market economies. Shallow interbank and money markets may impair the transmission from short-run policy interest rates to the structure of rates in the broader financial markets, while low levels of financial inclusion and widespread market power in the financial systems may impede transmission from the financial sector to the real economy.

Question 2. How do VAR methods seek to uncover the MTM?

Vector auto-regression (VAR) models are perhaps the most widely used methodology to analyze the MTM empirically. VARs used for monetary policy analysis entail the imposition of identifying restrictions on macro time-series data in order to trace the impact of innovations to interest rates or other monetary instruments (e.g., reserve money) on the evolution of a vector of macroeconomic aggregates, usually inflation, the output gap, exchange rates, and interest rates themselves but possibly also money and credit aggregates. Using VARs for monetary policy analysis, dating back to the seminal work of Sims (1980), confronts the researcher with formidable model specification and identification challenges—see for example Christiano and others (1999) for a discussion. To tackle these and respond to perceived weaknesses with the standard recursive VAR estimation, other methods have been applied to the advanced and emerging countries, including: generalized non-recursive SVARs (Sims and Zha, 1995; Kim and Roubini, 2000); sign restrictions (Uhlig, 2005); and factor-augmented VARs (Bernanke, Boivin, and Eliasz, 2005).

Most studies of the monetary transmission mechanism in LICs, such as those reviewed by Mishra, Montiel, and Spilimbergo (2012) and Mishra and Montiel (2013), still tend to rely on conventional SVAR models, with the majority identifying the monetary policy shock using recursive or block-recursive ordering or similar exclusion restrictions.

Question 3. What does the conventional VAR-based evidence tell us about the monetary transmission mechanisms in LICs?

The capsule summary of the extensive body of evidence surveyed by Mishra, Montiel, and Spilimbergo (2012) and Mishra and Montiel (2013) is that the MTM in low income countries is “relatively weak” and “less reliable” compared to advanced and emerging market economies. Movements in short-term policy rates are estimated to have a relatively weak pass-through on average to market rates, and movements in market rates have a similarly relatively weak estimated effect on aggregate demand and inflation. At the same time, these econometric estimates are less reliable in that they are bounded by much wider confidence intervals than is the case elsewhere, leaving considerable statistical uncertainty about the true MTM.

Question 4. Is the MTM really “weaker and less reliable” in LICs than elsewhere? What could affect the estimation when applying VAR methods to LICs?

There are two broad possible explanations for these findings:

  • Facts on the ground: As suggested earlier, formal financial markets in LICs are small and poorly arbitraged and exchange rates are often heavily managed so that the link between the short-term policy interest rates that central banks can control and the variables that matter for aggregate demand (e.g., longer-term interest rates, the exchange rate) may be weak or absent. Even the bank lending channel may be weak when the formal financial sector is small, financial frictions are severe, and the banking industry is characterized by imperfect competition.

  • Limitations of the method: The MTM is not in fact weak, but the VAR-based methods typically used to evaluate the MTM empirically are not capable of measuring its strength accurately in the research environment characteristic of LICs.

If the “facts on the ground” explanation is correct, this suggests that managing monetary policy successfully may be particularly difficult in low-income countries. Along with other features of the LIC environment, such as frequent large supply shocks, weak and uncertain transmission may exacerbate short-run interest rate volatility, and make it more difficult for policymakers to keep inflation within narrow bounds and stabilize activity in the face of demand shocks.

On the other hand, if the missing MTM mainly reflects methodological limitations, then the results of the VAR-based literature should be discounted by policymakers and researchers evaluating the strength and reliability of the MTM should seek empirical approaches that are more robust to the peculiar weaknesses of these methods in LIC-like environments.

Question 5. How do LIC-style environments affect VAR-based inference on the MTM?

In a forthcoming IMF working paper (Li and others, 2016), we employ a simulation (Monte Carlo) approach to discriminate between these two candidate explanations for the “missing MTM.” We generate data from a dynamic, stochastic general-equilibrium (DSGE) model of a small open economy in which a strong MTM is present. We first show that the monetary transmission mechanism can be cleanly and precisely identified in a favorable “advanced economy” research environment (think, for example, of Canada) and then seek to uncover this MTM using cleanly-identified SVAR methods, except for where we systematically introduce elements of the research environment that characteristically confront researchers and policymakers in LICs.

We find a surfeit of plausible explanations for the “missing MTM.” First, low interest rate and exchange rate elasticities in the goods market make the MTM harder to detect. Perhaps more interestingly, a low degree of smoothing in the interest rate reaction function, possibly a feature of less well-developed monetary policy frameworks, has similar effects on the strength of transmission and the ability of the VAR to detect. Of course, this feature could change rapidly with the policy regime.

What about the challenges of the research environment? We focus on three key features, both separately and in combination, although in each case we maintain the assumption that a strong and correctly-identified true MTM is present.

  • Short data samples: Macroeconomic regime shifts, such as changes in the exchange rate and/or monetary policy regimes or wide-ranging structural reforms in the financial sector often preclude the use of long historical data series, even when these exist. Compounded with data-collection limitations (in many low-income countries, for example, quarterly data on the real economy go back only about a decade) this means researchers are attempting to draw inference from relatively short spans of relevant historical data.

  • Measurement error: Macroeconomic variables display greater volatility at business-cycle frequencies in low-income countries than in higher-income countries, due in part to measurement errors in data-collection methods for output and inflation.

  • High-frequency supply shocks and conventional filtering methods: A defining characteristic of LICs is their greater exposure to high-frequency supply-side shocks, including climatic shocks to agricultural output. In these circumstances, the practice of proxying potential GDP with a smooth slow-moving trend may decant additional measurement error into output gaps computed by conventional filtering methods.

When present, these characteristics undermine inference in a consistent and intuitive manner (see Figure 1). Short data samples result in the estimated impulse response functions exhibiting a discernible attenuation over the early periods of the response (relative to the case where data are more abundant) and a widening of confidence bands around these estimates. Classical measurement error and errors arising from “over-smoothing” have qualitatively similar effects on inference, biasing the median estimated response functions towards zero and widening the error bounds relative to the baseline.

Figure 1.
Figure 1.

Estimating Impulse Responses in Hostile Empirical Settings

(“True” and SVAR-estimated impulse response functions of output to interest rate shocks)

Citation: IMF Research Bulletin 2016, 001; 10.5089/9781484316498.026.A003

Source: Li and others (forthcoming, IMF Working Paper).Note: The dotted blue line corresponds to the “true” impulse responses of the output gap to an interest rate shock, derived from the underlying DSGE model. The solid red lines correspond to the estimated median impulse response function and its 90 percent confidence interval derived from a recursively estimated SVAR model under alternative data characteristics.

In reality, these factors rarely occur in isolation. When allowed to play out together, the effects, while individually quite modest in scale, are exacerbated: the median estimated response of output to an interest rate shock falls very sharply relative to its true value while the confidence interval around this median widens. The VAR loses virtually all power to reject the false null of no transmission.

Question 6. Is robust measurement of the MTM even feasible in low-income countries?

The challenge for researchers and policymakers is whether the dominant VAR-based approaches can be bolstered, either by improved VAR methods or by complementary approaches that can help researchers triangulate on the true MTM. We have assumed that the VAR research can identify the monetary policy shocks, but the LIC environment this assumption—always a strong one—especially brave. The complex nature of policy regimes, often with an unclear and time-varying role for interest rates, money aggregates, and the exchange rate, is a deep challenge. Improved VAR-based methods may help here. For example, generalized non-recursive SVARs provide more flexibility in identifying structural shocks. Imposing sign restrictions is a more agnostic approach, restricting itself to defining only the signs on the impulse responses of certain shocks. Factor-augmented VARs allow for the use of simultaneous information contained in other variables. These can and should be complemented by approaches that are less reliant on crisp identification of macroeconomic data. Examples of alternative approaches include Berg and others (2013) who use a case-study approach to identify the impact of the large and coordinated monetary policy intervention by the central banks of Kenya, Uganda, Tanzania, and Rwanda in 2011, and Abuka and others (2015) who use loan-level data to assess the bank-lending channel in Uganda.

Question 7. What should monetary authorities do in low-income countries?

Even when inference is based on best practice VAR-based methods and bolstered by complementary approaches, uncertainty about the MTM is likely to continue to face central banks in low-income countries. Statistical uncertainty about VAR-based estimates is not, however, evidence of the absence of the MTM, and though it remains plausible that transmission in LICs is generally weaker and more uncertain than in other countries, neither is a reason for policy inaction. The notion that effective policy must be conditioned on precise and reliable quantitative understandings of the MTM is clearly a myth: the best must not stand as the enemy of the good. Monetary policymaking in all central banks, but arguably more so in LICs, entails a degree of “tatonnement,” of assessing the state of the economy, adjusting policy instruments accordingly, evaluating new data and the feedback evidence from the economy, and repeating this process. And for those LICs in particular that are in the process of reforming and revising their monetary policy frameworks, there is no real alternative to learning by doing, as well as effects of these reforms on the transmission mechanism itself (see IMF, 2015).

References

  • Abuka, Charles, Ronnie K. Alinda, Camelia Minoiu, Jose-Luis Peydro, and Andrea F. Presbitero. 2015. “Monetary Policy in a Developing Country: Loan Applications and Real Effects,IMF Working Paper 15/270. International Monetary Fund: Washington.

    • Search Google Scholar
    • Export Citation
  • Berg, Andrew, Luisa Charry, Rafael A. Portillo, and Jan Vlcek. 2013. “The Monetary Transmission Mechanism in the Tropics: A Narrative ApproachIMF Working Paper 13/197. International Monetary Fund: Washington.

    • Search Google Scholar
    • Export Citation
  • Bernanke, B., J. Boivin, and P. S. Eliasz. 2005. “Measuring the Effects of Monetary Policy: A Factor-Augmented Vector Autoregressive (FAVAR) ApproachQuarterly Journal of Economics, 120(1): 387422.

    • Search Google Scholar
    • Export Citation
  • Christiano, Lawrence J., Martin Eichenbaum, and Charles L. Evans. 1999. “Monetary Policy Shocks: What Have We Learned?” in John B. Taylor and Michael Woodford, eds., Handbook of Macroeconomics, Vol. 1A (Amsterdam: North Holland).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund. 2015. “Evolving Monetary Policy Frameworks in Low-income and Other Developing Countries,IMF Staff Report, October 2015.

    • Search Google Scholar
    • Export Citation
  • Kim, Soyoung and Nouriel Roubini. 2000. “Exchange Rate Anomalies in the Industrial Countries: A Solution with a Structural VAR Approach,Journal of Monetary Economics 45: 561586.

    • Search Google Scholar
    • Export Citation
  • Li, Bin Grace. 2008. “Evaluating Structural Vector Autoregression Models in Monetary Economies,University of Chicago, manuscript.

  • Li, Bin, Stephen O’Connell, Christopher Adam, Andrew Berg, and Peter Montiel. 2016. “VAR Meets DSGE: Uncovering the Monetary Transmission Mechanism in Low-Income Countries,IMF Working Paper, (forthcoming).

    • Search Google Scholar
    • Export Citation
  • Mishra, Prachi and Peter Montiel. 2013. “How Effective is Monetary Transmission in Low-Income Countries? A Survey of the Empirical EvidenceEconomic Systems, 37(2): 187216.

    • Search Google Scholar
    • Export Citation
  • Mishra, Prachi, Peter Montiel, Peter Pedroni, and Antonio Spilimbergo. 2014. “Monetary Policy and Bank Lending Rates in Low-Income Countries: Heterogeneous Panel Estimates,Journal of Development Economics, 111, November, 117131.

    • Search Google Scholar
    • Export Citation
  • Mishra, Prachi, Peter J. Montiel, and Antonio Spilimbergo, 2012. “Monetary Transmission in Low-Income Countries: Effectiveness and Policy Implications,IMF Economic Review, 60, 270302.

    • Search Google Scholar
    • Export Citation
  • Sims, Christopher A. 1980. “Macroeconomics and Reality,Econometrica, 48(1), 148.

  • Sims, Christopher A. and Tao Zha, 1995. “Does Monetary Policy Generate Recessions?: Using Less Aggregate Price Data to Identify Monetary Policy.Working paper, Yale University, CT.

    • Search Google Scholar
    • Export Citation
  • Uhlig, Harald, 2005. “What are the Effects of Monetary Policy on Output?: Results from an Agnostic Identification Procedure,Journal of Monetary Economics, 52, 381419.

    • Search Google Scholar
    • Export Citation
IMF Research Bulletin, March 2016
Author: International Monetary Fund. Research Dept.