Republic of Madagascar: Selected Issues
Author: International Monetary Fund. African Dept.
Publication Date:
30 Jan 2015
eISBN:
9781498353069
Language:
English
Keywords:
ISCR; CR; Madagascar; pass-through; price; oil price inflation; exchange rate movement; output gap; exchange rate adjustment channel; exchange rate appreciation; Exchange rates; Consumer price indexes; Vector autoregression; Consumer prices; Exchange rate pass-through
Download PDF (407.3 KB)

In recent years, the IMF has released a growing number of reports and other documents covering economic and financial developments and trends in member countries. Each report, prepared by a staff team after discussions with government officials, is published at the option of the member country.

Abstract

In recent years, the IMF has released a growing number of reports and other documents covering economic and financial developments and trends in member countries. Each report, prepared by a staff team after discussions with government officials, is published at the option of the member country.

Exchange Rate Pass-Through in Madagascar1

The exchange rate pass-through to domestic prices is estimated to be around -0.35 at its peak, which is similar to estimates for other Sub-Saharan African countries. There is also evidence that larger shocks to the exchange rate have a greater pass-through to prices relative to smaller shocks. This suggests that the authorities should allow the exchange rate to respond to shocks, rather than allowing imbalances to build, which will eventually lead to larger and more disruptive corrections.

A. Background

1. Exchange rate pass-through – the mechanism whereby fluctuations to the exchange rate impact domestic prices – has been an often-raised source of concern to low income countries (LICs). This can lead to a reluctance to allow the exchange rate to move in the face of shocks, which closes off an important economic adjustment channel. In Madagascar, inflation has been a source of economic vulnerability in the past – in the mid-1990s, consumer price inflation (CPI) peaked at close to 50 percent. And because high and volatile inflation can hit the poorest particularly hard, it is important that the authorities are mindful of these risks, while at the same time avoiding the build-up of harmful external imbalances. This chapter explores the degree of pass-through from changes in the nominal effective exchange rate (NEER) to the CPI in Madagascar.

2. Historically, both the change in CPI and NEER has been high and volatile in Madagascar (Table 1). The late 1980s and early 1990s witnessed particular instability in his regard, as Madagascar like other Sub-Saharan African (SSA) countries, unwound large internal and external imbalances. This period was also accompanied by weak economic growth. Exchange rate depreciations also followed political crises – such as in 2002 and 2009 – when confidence effects had a significant impact. Since 2003, however, inflation has been much more stable than previous years. A look at the data suggests there is some correlation between changes in NEER and CPI (Figure 1), especially during times when the variables experience significant spikes. The correlation coefficient between the two series is -0.44. But correlation does not necessarily imply causation. This study will test for casual relationships between the two variables, as well exploring the direction of causation. It will also explore whether other ‘omitted variables’ might be jointly determining innovations in these two series.

Table 1:

Summary Statistics: CPI and NEER*

(Year-over-year percent change)

article image

Positive = appreciation; 1985Q1-2014Q2

Figure 1:
Figure 1:

CPI and NEER

(Year-over-year percent change)

Citation: IMF Staff Country Reports 2015, 025; 10.5089/9781498353069.002.A004

Note: Positive value in the NEER=appreciation; some missing data in the early 1990s.Source: Malagasy authorities.

B. Literature and Methodology

3. There is an extensive literature on exchange rate pass-through, which can be divided into three main categories: i) structural macroeconomic models, including DSGEs, where economic theory determines model specification (see for example, Corsetti, Dedola and Leduc (2005)); ii) micro-econometric studies, which focus on individual markets or firms (see for example, Nucci and Pozzolo (2001)); and, iii) macro-econometric studies, predominantly using vector auto-regression (VAR) models, which are largely agnostic on the precise nature of the exchange rate and inflation shocks identified in the model and their transmission channels2. This study uses the VAR model approach. The advantage of the VAR method is that it is more robust to model specification errors, and the results are easier to interpret. The basic VAR structure is as follows:

Yt=c+BYt−n+CZt−n+ut(1)

where Yt is a vector of p endogenous variables; c is a vector of constants; B is a p × n matrix of coefficients for the endogenous variables; Zt is a vector of q exogenous variables; C is a q × n matrix of coefficients for the exogenous variables; and ut is vector of error terms, with a column length p.

4. The most basic potential VAR specification in this literature simply includes the NEER and CPI in the system. In equation 1, CPI and NEER would appear in the Yt, and Zt would be a vector of zeros. This technique is used by Razafimahefa (2012), surveying most SSA countries. In this paper, the average pass-through across all countries is estimated to be -0.2 after one-quarter, and -0.4 at four-quarters, after which the affect is small. There is significant heterogeneity between countries–in particular, those with more flexible exchange rate regimes and those with a higher income, tend to have a lower pass-through. Furthermore, the degree of pass-through has declined in the SSA region since the mid-1990s.

5. The disadvantage of this technique is that it is not able to control for reverse causality–specifically, that domestically generated inflation may be driving exchange rate shifts rather than vice-versa. This is particularly a problem if one believes that these relationships are not identically symmetric. Other studies attempt to control for demand and supply side shocks by including additional variables. McCarthy (1999) includes oil price inflation (as a proxy for supply shocks) and the output gap (as a proxy for demand shocks) in a study of nine OECD countries. Pass-through is estimated to be generally low, although it is larger for countries with a greater import share of GDP. Mwase (2006) uses a similar approach for estimating pass-through in Tanzania, and also includes money supply to help to control for demand shocks. By including import and producer prices as endogenous variables in the VAR, McCarthy (2000)3 explores how exchange innovations can be traced through the production distribution chain. Here exchange rate shocks first impact import prices, then affect producer prices, which in turn affect consumer prices.

6. There is also evidence that the size and direction of exchange rate movements can affect the degree of pass-through. Delatte and Lopez-Villavicencio (2012) find that exchange rate depreciations have a greater impact on consumer prices than exchange rate appreciations. They attribute this result to weak competition structures, whereby importers are able to increase profits rather than reduce prices as the exchange rate appreciates. In addition to finding evidence of asymmetric pass-through, Pollard and Coughlin (2003) find that the pass-through coefficient increases with the size of the exchange rate movement. For instance, a larger depreciation implies a greater share of the change will be passed on to prices. This may be because firms view larger changes in exchange rates as more permanent, thus requiring a more comprehensive change in pricing policy.

C. Results

7. Running the simple VAR with just CPI and NEER in the system generates a pass-through estimate of -0.34 for Madagascar after four quarters, after which the impact stabilizes (Table 2). The impulse response functions are shown in Figure A1. Further analysis (Table A1) provides evidence that shocks to NEER ‘Granger cause’ innovations in CPI. This suggests that the pass-through estimates can be viewed as not suffering from a reverse causality problem, although we cannot rule out that both series are driven by a common variable (omitted variable bias).

Table 2:

Cumulative Pass-Through Estimate to CPI from a Shock to NEER

article image
Figure A1:
Figure A1:

Impulse Response for Simple VAR Specification

Citation: IMF Staff Country Reports 2015, 025; 10.5089/9781498353069.002.A004

Note: Lag length of 4. Unit root tests suggest that all variables are stationary.

8. In order to control for this (following McCarthy (1999)), the output gap4 and international oil prices5 are introduced into the model. These results show a much higher pass-through rate, both contemporaneously (-0.14) and after 4 quarters (-0.48). However, the results here are less robust. The impulse response (Figure A2) shows that the change in CPI to changes in NEER is not statistically significant. Furthermore, a Granger causality test suggests that neither NEER (nor the output gap) causes changes in CPI. On further consideration, there are two important reasons why the specification used by McCarthy (1999) for advanced economies, may not be appropriate for Madagascar. First, the output gap is a difficult concept to measure in LICs, especially using a HP-filter, not least because supply side shocks are more important in determining output fluctuations than in advanced economies. Second, the authorities in Madagascar have in the past prevented changes in oil prices from being passed onto consumer prices (at often significant fiscal expense). These two factors suggest that the output gap and international oil prices are poor proxies for demand and supply shocks in Madagascar.

Figure A2:
Figure A2:

Impulse Response for VAR Specification with Output Gap and Oil Prices

Citation: IMF Staff Country Reports 2015, 025; 10.5089/9781498353069.002.A004

Note: Lag length of 4. Unit root tests suggest that all variables are stationary.

9. An alternative specification, better tailored for Madagascar, might be to include money supply and the rice harvest as better proxies for demand and supply shocks, respectively. Both narrow money (M0) and broad money (M3) are separately included in the VAR specification, although both give similar results. The rice harvest is included as an exogenous variable, while money supply is considered as endogenous to the system. The pass-through estimate is negligible on impact, rising to around -0.25 after two quarters, peaking at around -0.35 after 1 year. The impulse response functions are shown in Figure A3 and A4, and these results are robust to changing the Cholesky factorization. Both specifications suggest that money supply and the NEER Granger cause changes to CPI.

Figure A3:
Figure A3:

Impulse Response for VAR Specification with M0 and Rice Harvest

Citation: IMF Staff Country Reports 2015, 025; 10.5089/9781498353069.002.A004

Note: Lag length of 4. Unit root tests suggest that all variables are stationary.
Figure A4:
Figure A4:

Impulse Response for VAR pecification with M3 and Rice Harvest

Citation: IMF Staff Country Reports 2015, 025; 10.5089/9781498353069.002.A004

Note: Lag length of 4. Unit root tests suggest that all variables are stationary.

10. There is evidence that the size and direction of exchange rate movements affects the pass-through estimate. When focusing on exchange rate appreciations, the pass-through estimate is statistically insignificant, suggesting that prices do not fall in such circumstances. The pass-through estimate for depreciations is unchanged at -0.35 after four quarters. Therefore there is evidence of asymmetric pass-through depending on the direction of the exchange rate movement. Caution, however, is needed when interpreting this result, as there are only a few events of appreciation in the sample, suggesting there may be a small sample bias. Perhaps it is more interesting to consider how the size of the depreciation impacts the pass-through coefficient. Dividing the sample into depreciations of above and below 2 percent in a quarter (around the sample median) gives very different estimates. ‘Small’ depreciations have a statistically insignificant impact on consumer prices, while ‘large’ depreciations have a higher pass-through estimate. The pass-through rate in this latter specification is -0.42 after 4 quarters, significantly higher than the estimate for the entire sample.

D. Conclusions

11. Exchange rate pass-through in Madagascar is estimated to be around -0.35, under the preferred model specification (including money supply and the rice harvest). This is similar to estimates for other SSA countries, especially those with a floating exchange rate regime. This suggests that Madagascar is not especially vulnerable to pass-through relative to its peers. Furthermore, pass-though is transmitted gradually over four-quarters, suggesting that the authorities have some time to implement mitigating policy measure, if necessary.

12. The size and direction of exchange rate movements also matter. There is some evidence that depreciations are more likely to be transferred to consumer prices than appreciations. But perhaps more importantly, larger exchange rate movements are more likely to be pass-through to prices than smaller shocks. This suggests that gradual and frequent movements in the exchange rate are preferable to large infrequent movements. The policy implication therefore is that the authorities should be cautious when adopting a policy of preventing exchange rate movements, especially if it’s not clear that these are temporary. This strategy is preferable to closing off the exchange rate adjustment channel, which is important for absorbing external shocks and preventing the build-up of imbalances.

E. Annex

Table A1:

Granger Causality Test (simple specification)

VAR Granger Causality/Block Exogeneity Wald Tests

Sample: 1984Q4 2014Q2 IF D_CPI>-0.8

Included observations: 106

article image
Table A2:

Granger Causality Test (M0 and rice harvest)

VAR Granger Causality/Block Exogeneity Wald Tests

Date: 10/31/14 Time: 11:21

Sample: 1984Q4 2014Q2 IF D_CPI>-0.8

Included observations: 106

article image
Table A3:

Granger Causality Test (M3 and rice harvest)

VAR Granger Causality/Block Exogeneity Wald Tests

Date: 10/31/14 Time: 11:27

Sample: 1984Q4 2014Q2 IF D_CPI>-0.8

Included observations: 106

article image

References

  • Corsetti, G., Dedola, L. and Leduc, S. (2005) “DSGE models of high exchange-rate volatility and low pass-through” Board of Governors of the Federal Reserve System, International Finance Discussion Papers 845.

    • Search Google Scholar
    • Export Citation

  • Delatte, A. and Lopez-Villavicencio, A. (2012) “Asymmetric exchange rate pass-through: Evidence from major countries” Journal of Macroeconomics.

    • Search Google Scholar
    • Export Citation

  • Duma, N. (2008) “Pass-Through of External Shocks to Inflation in Sri Lanka” IMF Working Paper No. 08/78.

  • McCarthy, J. (1999) “Pass-Through of Exchange Rates and Import Prices to Domestic Inflation in some Industrialized Economies” BIS Working paper No 79.

    • Search Google Scholar
    • Export Citation

  • Mwase, N (2006) “An Empirical Investigation of the Exchange Rate Pass-Through to Inflation in Tanzania” IMF working paper 150.

  • Nucci, F. and Pozzolo, A. (2001) “Investment and the Exchange Rate: An Analysis with Firm-Level Panel Data” European Economic Review.

    • Search Google Scholar
    • Export Citation

  • Pollard, P. and Coughlin, C.> (2003) “Size Matters: Asymmetric Exchange Rate Pass-Through At The Industry Level” Federal Reserve Bank of St Louis Working Paper 2003-029C.

    • Search Google Scholar
    • Export Citation

  • Razafimahefa, I. (2012) “Exchange Rate Pass-Through in Sub-Saharan African Economies and its Determinants” IMF Working Paper No. 12/141.

    • Search Google Scholar
    • Export Citation

1

Prepared by Alex Pienkowski.

2

For example, a structural model might differentiate between productivity and nominal shocks both in Madagascar and abroad, which would have theoretically different implications on inflation and exchange rates. The VAR approach can, at best, approximate these relationships, and so should be viewed as identifying ‘reduced form’ relationships.

3

And adopted by Duma (2008) in the case of Sri Lanka.

4

Following Mwase (2006), the quarterly data is interpolated from annual observations, and an HP-filter is used to construct the output gap estimate.

5

International oil prices are included as an exogenous variable; including the output gap as an exogenous or endogenous variable does not materially change the pass-though estimates (the latter specification is reported).

Other IMF Content

Other Publishers

Cover IMF Staff Country Reports
Republic of Madagascar: Selected Issues
Author: International Monetary Fund. African Dept.
Volume/Issue:
Volume 2015: Issue 025
Publisher:
International Monetary Fund
ISBN:
9781498353069
ISSN:
1934-7685
Pages:
64
DOI:
https://doi.org/10.5089/9781498353069.002

CPI and NEER

(Year-over-year percent change)

Impulse Response for Simple VAR Specification

Impulse Response for VAR Specification with Output Gap and Oil Prices

Impulse Response for VAR Specification with M0 and Rice Harvest

Impulse Response for VAR pecification with M3 and Rice Harvest