Republic of Kazakhstan: Selected Issues
Author: International Monetary Fund. Middle East and Central Asia Dept.
Publication Date:
09 May 2017
eISBN:
9781475598759
Language:
English
Keywords:
ISCR; CR; price; Kazakhstan; exchange rate; oil price; production function approach; interest rate rule; trend growth; D. production function; import price shock; Oil prices; Exchange rates; Real interest rates; Exchange rate pass-through; Fiscal stance; Global; Eastern Europe; monetary policy decision; equilibrium pass-through effect; price variable; price chain; price volatility; Inflation; Import prices; Consumer prices; interaction term
Download PDF (642.3 KB)

Selected Issues

Abstract

Selected Issues

Exchange Rate Pass-Through in Kazakhstan: Empirical Findings and Implications for Inflation Targeting1

This paper examines the exchange rate pass-through (ERPT) in Kazakhstan in the context of inflation targeting. Knowledge of the timing and magnitude of ERPT is important for the conduct of monetary policy. Using both single equation and system methods, the pass-through to consumer prices is estimated in the range of 20- 30 percent. This is close to findings for other developing and emerging market economies and is consistent with Kazakhstan’s own recent experience. Empirical evidence is accumulating that monetary policy can influence ERPT; putting a credible monetary policy framework in place—supported by effective communications—may anchor expectations and be associated with lower pass-through. Reducing dollarization and introducing more competition in the domestic product markets could also help reduce ERPT.

A. Introduction

1. Kazakhstan has started a transition to inflation targeting. On August 20, 2015, the National Bank of Kazakhstan (NBK) released a joint statement with the government laying out the key elements of a transition to inflation targeting (IT) over the medium term.2 Among other things, the new monetary policy framework implies a change in the exchange rate regime toward greater exchange rate flexibility. Consistent with this change, the NBK announced that it would no longer intervene in the foreign exchange market, except to prevent excessive volatility that could have a destabilizing effect on the financial system.

2. The decision to abandon the exchange rate band and allow the tenge float was followed by in a sharp depreciation. At the close of the morning session of the Kazakhstan Stock Exchange (KASE) on August 20, the tenge traded at 255 per US dollar, up from KZT 188 per USD a day earlier. The market stabilized shortly thereafter but in mid-September, tensions reemerged in an environment of low trading volumes in the forex market (National Bank of Kazakhstan, 2015a). The NBK resumed interventions to contain volatility. However, in Q4 of 2015, against the background of falling oil prices and weakening of the Russian ruble, the pressure on the tenge intensified. The NBK continued to intervene but in early November it withdrew again from the market to preserve foreign reserve buffers (National Bank of Kazakhstan, 2015b). The exchange rate continued to depreciate and eventually stabilized at around KZT 340 per USD.

3. Inflation accelerated noticeably. Consumer prices increased significantly in October and November 2015, and inflation reached 13.6 percent at the end of the year (compared to 4.4 percent in September). The price increase encompassed all major groups of goods and services, with prices of non-food items, which are primarily imported, growing by the highest rate (National Bank of Kazakhstan, 2015a). 2016 saw a further acceleration of inflation which peaked at 17.7 in July – almost a year after the first depreciation episode.

4. Gaining a better understanding of the link between exchange rates and inflation is key for the conduct of monetary policy. The recent exchange rate developments and the subsequent inflation dynamics bring to the fore the issue of the impact of depreciation on domestic prices, usually referred to as exchange rate pass-through (ERPT). ERPT is important for at least two reasons.

  • First, inflation forecasts are one of the main building blocks of IT and the exchange rate is an essential ingredient in the forecasts, especially in small open economies. In this regard, both the short-term and the long-term effects are important as exchange rate movements are usually transmitted to domestic prices with a lag. Many central banks, including well-established inflation targeters such as the Bank of England and the ECB, have explicitly reflected the potential inflationary effects of weaker exchange rates into their monetary policy decisions.

  • Second, ERPT has implications for the effectiveness of monetary policy. This is related to the role of the nominal exchange rate as a shock absorber (Edwards, 2006). The impact of a nominal depreciation on the real exchange rate depends on the ERPT, and the real exchange rate is an important determinant of a country’s external position. Low ERPT implies smaller expenditure-switching effect and greater effectiveness of monetary policy in stimulating the domestic economy (Campa and Goldberg, 2005). At the same time, as argued by Edwards (2006), it is important to distinguish between the pass-through into the prices of tradables and non-tradables; a high ERPT for non-tradables would reduce the effectiveness of monetary policy, while a high ERPT for tradables would increase it.

5. Channels through which the exchange rate influences domestic prices can be described following the distribution chain. Typically, the process comprises two stages. In the first stage, exchange rate fluctuations are reflected in import prices and in the second stage, they are transmitted to consumer prices, either directly through the imported component of the CPI, or indirectly through producer prices. From the perspective of a central bank that targets inflation, the ultimate effect of exchange rate movements on the CPI is of primary interest.

6. There is a vast literature on the subject, including a number of survey articles. For example, a recent paper by Aron and others (2014) contains a comprehensive review of both the theoretical underpinnings of ERPT and empirical results. Overall, nearly all studies have found evidence of incomplete ERPT. This is related to the firms’ pricing strategies. The literature distinguishes two main types of such strategies: (i) producer currency pricing (PCP) where prices are set in the exporter’s currency and (ii) local currency pricing (LCP). Under PCP, since exporters’ prices are broadly unchanged, the ERPT to import prices would be expected to be more or less complete. In practice, however, the local prices usually do not react one to one to exchange rate changes, yielding support to the LCP hypothesis. Factors like degree of competition/segmentation of the domestic market and substitutability between domestically-produced and imported goods are likely to matter as well. The level of dollarization may also play a role. Some studies suggest that dollarization appears to increase ERPT, supporting the claim that “fear of floating” is a more serious issue in dollarized economies (Reinhart and others, 2014). Gonzales (2003), on the other hand, finds no relationship between the degree of dollarization and the pass-through to domestic inflation for thirteen Latin American countries between 1980 and 2000.

7. Many studies have documented a decline in the degree of ERPT over time and have proposed various explanations. These include increased integration and changes in the market structure and the structure of consumer expenditures as reflected in the weights of the CPI components (Aron and others, 2014).3 Another hypothesis, put forward by Taylor (2000), is that lower ERPT is due to the more benign inflation environment that many countries have achieved. In such an environment, firms have less pricing power, which prevents them from passing increased costs on to the consumer. Enhanced credibility of monetary policy, including through introduction of IT, has likely contributed to lower and more stable inflation.

8. As regards empirical strategies, two main approaches have been used to estimate ERPT. Much of the empirical work has been carried out using single equation techniques but more recent contributions tend to rely more on system models, such as VARs and VECMs. Both approaches have advantages and disadvantages. Single equation methods can be linked to simple theoretical propositions (e.g. variations of the law of one price) and have more flexibility in terms of econometric specifications, as they allow for including asymmetric responses, structural breaks, and other types of non-linearities. On the other hand, vector based models recognize the endogenous nature of exchange rates and prices. From a policy making perspective, it is perhaps best to estimate the ERPT using an array of different techniques and compare the results, keeping in mind the potential biases of the respective methods.

9. Estimates of ERPT have varied significantly, depending on the type of country, sample period and method. For developing and emerging market countries, Frankel and others (2012) find that the ERPT to CPI is 34 percent after one year. Choudhri and Hakura (2006) have a broadly similar estimate – 27 percent after 20 quarters for the entire sample. Billmeier and Bonato (2002) estimate a pass-through of about 33 percent for the retail price index in Croatia. Beirne and Bijsterbosch (2009) examine the degree of ERPT in nine Central and Eastern European countries and report an average ERPT of 60 percent using a cointegrated VAR and 50 percent based on a VAR in differences.

10. This paper looks into the effects of exchange rate changes on price developments in Kazakhstan. Its primary objective is to assess the pass-through to consumer prices given that it is of main concern for the conduct of monetary policy. However, occasionally results for import prices are reported as well since imports prices provide a more immediate link to exchange rate developments. Following the literature, we use both single equation specifications and vector-based methods and check the sensitivity of estimates.

B. Data Description and Properties

11. The empirical analysis focuses on six main variables typically considered in the ERPT literature. Specifically, we use monthly data for consumer prices (CPI), producer prices (PPI), import prices, industrial production, nominal effective exchange rate and external prices in the period January 2000 - June 2016. The sources of data are indicated in the Appendix.

12. Unit root tests indicate non-stationarity of the data in levels. A visual inspection of the original price series and the exchange rate (in logarithms) reveals a close co-movement, especially of import prices and the PPI which appear highly correlated (Figure 1). However, the data in levels are likely non-stationary, so we perform unit root tests to formally check for stationarity. Since the series are upward trending, we test the null that the data follow a random walk with a drift vs. the alternative of a stationary process around a linear trend. Both the Augmented Dickey-Fuller test and the Phillips-Perron test (which accounts for potential serial correlation and heteroscedasticity) suggest that for the level data, the null for unit root cannot be rejected (Table 1). The same tests applied to first differences strongly reject the null, indicating that all series are likely I(1).

Figure 1.
Figure 1.

Exchange Rate and Various Price Indices (logarithmic scale)

Citation: IMF Staff Country Reports 2017, 109; 10.5089/9781475598759.002.A003

Source: IMF staff calculations
Table 1.

Kazakhstan: Unit Root Tests

article image
Source: IMF staff calculations

13. Inflation appears to be significantly correlated with exchange rate movements. As an initial step in the analysis, we look at the bilateral correlations of the main variable of interest – inflation – with the changes in the exchange rate. For the whole sample, the correlation coefficients between the change in log CPI and log NEER and its first three lags are 0.27, 0.46, 0.37 and 0.15, respectively. However, the strength of association between the two variables has changed significantly over time (Figure 2); it has become much tighter in the more recent period which is characterized with higher exchange rate and price volatility.

Figure 2.
Figure 2.

Kazakhstan: Correlations between Inflation and Exchange Rate Changes

Citation: IMF Staff Country Reports 2017, 109; 10.5089/9781475598759.002.A003

Source: IMF staff calculations

C. Single Equation Models

14. Earlier work on ERPT has focused on import prices and has relied exclusively on single equation methods. For instance, one of the widely cited studies by Campa and Goldberg (2005) is based on estimating an equation for the differenced log import prices as a function of the exchange rate, foreign production costs and real GDP and their first four lags. In this setup, the short-run relationship between import prices and exchange rates is captured by the contemporaneous coefficient, while the long-run elasticity is given by the sum of the coefficients on the contemporaneous exchange rate and its lags. Mihaljek and Klau (2008) estimate a similar model for the CPI and include lags of the dependent variable among the regressors, as well as the equilibrium real exchange rate gap as a control variable. 4 Similar to Campa and Goldberg (2007), quarterly data are employed and up to four lags of the explanatory variables are included in the estimation.

15. With monthly series, choosing the appropriate lag length is less straightforward. Determining the number of lags involves a trade-off between capturing delayed effects and sacrificing degrees of freedom. Generally, information criteria can guide selection but without imposing any constraints, the number of possible combinations grows rapidly with the number of lags and often different information criteria point to different specifications. To make the process more manageable, we fix the number of exchange rate lags at six5 and vary the number of lags of industrial output and foreign prices from six to one. Thus, the family of models we estimate is:

Δcpit=α+Σi=06Δet−i+Σi=0kΔyt−i+Σi=0kΔp*t−i+ɛt,

where Δcpit, Δet, Δyt and Δp *t stand for the first difference of the logarithms of CPI, NEER, industrial output and foreign prices, respectively. The Akaike Information Criterion suggests using six lags (k=6) for output and foreign prices, while the Bayesian Information Criterion (BIC) selects 1 lag (k=1). The estimated coefficients to the exchange rate and its lags are reported in Table 2 (the coefficients on the industrial output and foreign prices and the constant are not reported for brevity).

Table 2.

Kazakhstan: Estimated Coefficients to NEER and Its Lags

article image
t statistics in parentheses* p < 0.05, ** p < 0.01, *** p < 0.001Source: IMF staff calculations

16. The different specifications differ little in terms of estimated elasticities. In both versions (with one and six lags), the point estimate of the short-term elasticity of inflation to exchange rate depreciation is very small – about 2 percent and the coefficient is not statistically significant. The coefficients on the first and second lag, however, are highly significant, implying that it takes at least a month for exchange rate changes to start having an impact on consumer prices. The sum of all coefficients on the exchange rate is about 0.155 in Model A and 0.141 in Model B; including more lags of the control variables has little effect on the longer-term elasticities.

17. In addition, we test for asymmetry and non-linearities. Models C through E take the baseline specification of model A and include additional variables to test whether the introduction of exchange rate flexibility in the context of IT has had an impact on the ERPT and whether there are asymmetries or threshold effects. Specifically, “IT” is a binary variable taking the value of 1 from August 2015; “Depreciation” is also a binary variable equal to 1 if the exchange rate depreciated in the corresponding month and 0 otherwise, and “Large depreciation” is an interaction term of the exchange rate change and a dummy variable indicating whether depreciation was more than 3 percent in a given month (1 percent and 5 percent were also tried without qualitatively changing the results). None of these additional variables seems to have a statistically significant impact on inflation.

18. Robustness checks indicate that estimates are not very sensitive to alternative specifications. For example, replacing industrial output with an output gap proxy (calculated using the HP filter), as in Mihaljek and Klau (2008), has a marginal impact on the results. Finally, applying the baseline specification (Model A) to import prices instead of the CPI yields higher elasticities – both in the short and the long run. The estimated contemporaneous coefficient on the exchange rate is 0.065 and the sum of all coefficients on this variable is 0.247, implying a faster and larger pass-through to import prices.

D. System Methods

19. The main advantage of system methods is that they treat variables as endogenous. While univariate methods provide some flexibility as discussed above, they do not account for the potential endogeneity of the exchange rate and the price variables. This is why many studies, especially more recent ones, have resorted to system methods to make inferences about the effect of the exchange rate movements on domestic prices. The two preferred frameworks are cointegrated VAR and VAR in differences. An example of the latter is Winkelried (2014) who documents a decline in ERPT in Peru after the adoption of a fully-fledged IT regime. He considers a 6-variable SVAR which follows the price chain, with an output variable reflecting the state of the economy.6 Identification is based on the Cholesky decomposition of the covariance matrix. Since the VAR is formulated in differences, the relevant information on ERPT is provided by the cumulative impulse responses.

20. A stationary VAR model with recursive identification can be used to examine the ERPT in Kazakhstan. Following the literature, we consider the vector (p*, e, pm, y, ppi, cpi), where the variables denote log differences of foreign prices, exchange rate, import prices, industrial output, producer and consumer prices (as defined above) in this order. The specification differs slightly from Winkelried (2014) as in his model output is placed second and a constraint on the demand shock is imposed to achieve identification. In general, the ordering of variables is important in recursive identification schemes, which essentially assume that the error term in each equation is uncorrelated with the error term in the previous equation. In our specific example, this amounts to assuming that foreign supply, exchange rate and import price shocks have a contemporaneous effect on output but a demand shock does not affect the first three variables instantaneously. This seems a plausible hypothesis in the case of Kazakhstan. Moreover, the results are not sensitive to the position of output; the impulse responses of the price variables to an exchange rate shock look very similar in alternative specifications where output is placed second or third. For the rest of the variables, the recursive structure is justified on economic grounds as the causal chain logically runs from foreign prices to exchange rate, import prices, producer prices and finally to consumer prices.

21. Formal lag selection criteria are used to determine the optimal number of lags. The Akaike, Hannan-Quinn and Schwarz information criteria all suggest only one lag, whereas the LR criterion points to five lags. Estimating the VAR with one lag, however, is problematic as there is significant residual autocorrelation which violates the assumption of independence. It is evident from an inspection of the correlograms and is also confirmed by formal testing. Autocorrelation is removed when five lags are selected, so we estimate a VAR (5) model.

22. Results suggest that the pass-through of exchange rate changes to CPI is around 30 percent. Figure 3 (left panel) shows the orthogonalized impulse response of consumer prices to one standard deviation shock in the exchange rate in the baseline specification. The response is significant and peaks 1-2 months after the depreciation. Since the VAR model is formulated in differences, the quantity of primary interest is the accumulated impulse response (Figure 3, right panel). To obtain the ERPT, impulse responses need to be transformed, so that they correspond to a one percent shock in the exchange rate. Calculations suggest that the contemporaneous response is very small – about 2 percent; the accumulated effect increases to about 25 percent after 3 months and to 30 percent after 6 months. The reaction of prices to an ER shock, as captured by the impulse response functions, suggests some overshooting, so after the first six months the effect of the depreciation diminishes somewhat and stabilizes at around 27 percent in the long run.

Figure 3.
Figure 3.

Kazakhstan: Impulse Response of CPI to One SD of NEER

Citation: IMF Staff Country Reports 2017, 109; 10.5089/9781475598759.002.A003

Source: IMF staff calculations

23. For comparison, import prices respond more strongly to depreciation – the ERPT after 3 months is 30 percent and after 6 months it is 55 percent which is also the long-term effect (Figure 4). However, the confidence bands are much wider around these estimates. The higher pass-through to import prices is not surprising since part of the depreciation effect is dampened in the distribution phase. 7

Figure 4.
Figure 4.

Impulse Response of Import Prices to One SD of NEER

Citation: IMF Staff Country Reports 2017, 109; 10.5089/9781475598759.002.A003

Source: IMF staff calculations

24. As noted above, the estimates of the ERPT are quite robust with respect to reordering of the industrial production variable. As another robustness check, we estimate the model with different numbers of lags and an output gap variable, as well as with seasonally unadjusted index of industrial production.8 The results remain broadly unchanged (see Figure A1 in the Annex). Finally, replacing the nominal effective exchange rate with the KZT/USD exchange rate and re-estimating the model with 2 lags produces an ERPT of 28 percent in the long run – very close to the baseline estimate (Figure A2 in the Annex). The latter result is important since it is the exchange rate to the USD that is observed directly and often forms the basis for pricing decisions in domestic currency.

25. The VAR in difference provides useful insights but omits information on a possible long-term relationship between the variables in the model. Testing for such a relationship amounts to testing for cointegration – a linear combination of I(1) variables that is stationary. A number of studies (see Coricelli and others (2006), and Beirne and Bijsterbosch (2009), among others) have addressed the ERPT problem within the cointegrated VAR framework. The cointegration approach has been tried for Kazakhstan as well but two issues have emerged – one theoretical and one related to the estimates – that prevent us from presenting it as the preferred method for ERPT assessment.

26. The theoretical challenge pertains to the interpretation of the coefficients in the cointegrating relationship. It is customary when variables are in logarithms to interpret these coefficients as long run elasticities, especially when there is a single cointegrating equation. If there are more than one cointegrating vectors, which is often the case with larger systems, the economic meaning of the coefficients is less straightforward since any linear combination of such vectors is also a cointegrating vector. This leads to indeterminacy and some authors (e.g. Beirne and Bijsterbosch, 2009) have proposed to take the first cointegrating vector which has the highest eigenvalue. However, as pointed out by Masten (2004), the correct identification of the equilibrium pass-through effect is more involving and in many cases it is not feasible unless restrictions are placed on the reactions of some of the variables. According to his Proposition 1, the equilibrium ERPT is identified if and only if the cointegrating rank of the system is equal to 1 plus the number of variables with non-zero coefficients in the vector of long-run responses. This conclusion is based on an earlier result by Johansen (2002) on the interpretation of cointegrating coefficients.

Table 3.

Cointegrating Vectors and Adjustment Coefficients

article image
Source: IMF staff calculations

27. In the case of Kazakhstan, the Johansen test suggests two cointegrating relationships for the baseline specification. The baseline specification includes the same variables as the stationary VAR, 5 lags, a linear trend in the level data and cointegrating equations that are stationary around a nonzero mean. The same result obtains for a specification that allows for linear trends in the cointegrating equations based on the maximum eigenvalue statistic. Estimation results for the unrestricted model with the Johansen identification scheme are shown in Table 3. The first cointegration equation has the form:

cpi=1.277pm−0.515y+0.745e+1.527p*

and all coefficients, except the one on the output variable, are significant at the 5 percent level. The estimated coefficient on the nominal effective exchange rate is 0.75 but as discussed above, it does not necessarily measure the pass-through; restrictions need to be imposed to determine the elasticity of prices to changes in the exchange rate. Such restrictions could either come from theory or from the data. Theory, however, does not seem to suggest specific values for some of the long run changes as these would be expected to vary across countries. We can constrain some of the elements of the cointegrating vectors which are not statistically significant to be equal to zero, e.g. industrial output in both equations and import prices in the second one, but this would still not be enough to achieve identification.

28. The empirical issue is related to the estimates of the adjustment coefficients which are in most cases close to zero and statistically insignificant. This indicates weak exogeneity with respect to the cointegrating parameters. While it is intuitive that foreign prices would be weakly exogenous, domestic prices would be expected to adjust in a significant way to deviations from equilibrium. This, however, is not supported by the data; the only variable that adjusts significantly in a consistent manner is the exchange rate. Further, the results are rather sensitive to the model specification. For example, replacing the seasonally adjusted output variable with the unadjusted one (also with 5 lags to remove residual autocorrelation), yields a coefficient of 0.54 on the exchange rate in the first cointegration equation which is not statistically significant. Using the USD exchange rate (and 3 lags) produces an estimate of 0.34 and highly statistically significant, which is more in line with the VAR in differences results.

E. Conclusions and Policy Implications

29. Results from the empirical analysis of ERPT in Kazakhstan reveal a moderate impact of exchange rate changes on domestic inflation. A comparison across the different estimation methods employed suggests that the contemporaneous response of prices is rather small but the effect increases significantly one and two months after the exchange rate shock. Based on the estimates of the VAR model in differences, which is our preferred analytical tool, a 10 percent depreciation of the tenge would be expected to trigger an increase in consumer prices on the order of 3 percent. This is close to findings for other developing and emerging market economies and is consistent with Kazakhstan’s own recent experience.

30. The timing and magnitude of ERPT has potential implications for monetary policy. In the context of inflation targeting, the inflationary effect of a weakening of the domestic currency has to be factored in when decisions about the policy rate are made. The degree of ERPT is also important for the speed of external adjustment following a shock under a flexible exchange rate regime. If the domestic currency depreciates and the pass-through to import and producer prices is high, expenditure switches toward locally produced goods as imports become relatively expensive. If then the pass-through to consumer prices is low, the nominal depreciation would also imply real depreciation and competitiveness would improve, which in turn would abate pressures on the exchange rate. Empirical evidence is accumulating that monetary policy itself can influence ERPT; putting a credible monetary policy framework in place anchors expectations and tends to be associated with lower pass-through. In this regard, besides incorporating the information on ERPT in its monetary policy decisions, the NBK would benefit from a clear communication to the public of the anticipated effects on prices following an exchange rate shock.

31. Reducing dollarization and introducing more competition in the domestic product markets could help reduce ERPT. Financial dollarization amplifies the effects of exchange rate shocks and could be associated with higher pass-through to consumer prices. Therefore, efforts to de-dollarize, including by deepening the financial markets and encouraging the development of hedging instruments is key. On the structural front, greater competition is more likely to be associated with variable markups which firms can adjust downwards in case of depreciation in order to remain in the business. In this regard, Kazakhstan can benefit from a faster implementation of the government’s privatization program and reforms aimed at reducing the role of state monopolies and promoting private entrepreneurship.

References

  • Aron, J., Macdonald, R. and J. Muellbauer (2014), “Exchange Rate Pass-Through in Developing and Emerging Markets: A Survey of Conceptual, Methodological and Policy Issues, and Selected Empirical Findings”, Journal of Development Studies, Vol. 50, No. 1, pp. 101–143

    • Search Google Scholar
    • Export Citation

  • Beirne, J. and M. Bijsterbosch (2009), “Exchange Rate Pass-Through in Central and Eastern European Member States”, ECB Working Paper, No. 1120

    • Search Google Scholar
    • Export Citation

  • Berger, D., Faust, J., Rogers, J. and K. Steverson (2009), “Border Prices and Retail Prices”, Board of Governors of the Federal Reserve System, International Finance Discussion Papers, No. 972

    • Search Google Scholar
    • Export Citation

  • Choudhri, E. and D. Hakura (2006), “ERPT to domestic prices: Does the inflationary environment matter?”, Journal of International Money and Finance, Vol. 25, pp. 614–639

    • Search Google Scholar
    • Export Citation

  • Billmeier, A. and Bonato, L. (2002), “Exchange Rate Pass-Through and Monetary Policy in Croatia”, IMF Working Paper, No. 02/109

  • Campa, J. and L. Goldberg (2005), “Exchange Rate Pass-Through into Import Prices”, Review of Economics and Statistics, 87, pp. 379–390

    • Search Google Scholar
    • Export Citation

  • Coricelli, F., Jazbec, B. and I. Masten (2006), “Exchange Rate Pass-Through in Acceding Countries: Empirical Analysis and Policy Implications”, Journal of Banking and Finance, Vol. 30, No. 5, pp. 1375–1391

    • Search Google Scholar
    • Export Citation

  • Edwards, S. (2006), “The Relationship Between Exchange Rates and Inflation Targeting Revisited”, NBER Working Paper, No. 12163

  • Frankel, J., Parsley, D. and S. Wei (2012), “Slow Pass-Through around the World: A New Import for Developing Countries?”, Open Economies Review, Vol. 23, No. 2, pp. 213–251

    • Search Google Scholar
    • Export Citation

  • Gonzales, J.A. (2003), “Exchange Rate Pass-Through and Partial Dollarization: Is There a Link?”, in Latin American Economic Reforms: The Second Stage (Gonzales, J.A., Corbo, V., Krueger, A. and A Tornell, editors)

    • Search Google Scholar
    • Export Citation

  • Johansen, S. (2002), “The Interpretation of Cointegrating Coefficients in the Cointegrated Vector Autoregressive Model”, Preprint No. 14, Department of Theoretical Statistics, University of Copenhagen.

    • Search Google Scholar
    • Export Citation

  • Lutkepohl, H. (2011), “Vector Autoregressive Models”, EUI Working Papers, ECO2011/30

  • Masten, I. (2004), “Identification of Exchange Rate Pass-Through Effect in Cointegrated VAR: An Application to New EU Member Countries”, preprint available at http://webv1ef.ef.uni-lj.si/dokumenti/wp/clanek161.pdf.pdf

    • Search Google Scholar
    • Export Citation

  • Mihaljek, D. and M. Klau (2008), “Exchange Rate Pass-Through in Emerging Market Economies: What has Changed and Why?”, BIS Papers, 35, pp. 103–130

    • Search Google Scholar
    • Export Citation

  • National Bank of Kazakhstan (2015a), “Inflation Report”, The Third Quarter of 2015, Almaty, Kazakhstan

  • National Bank of Kazakhstan (2015b), “Inflation Report”, The Fourth Quarter of 2015, Almaty, Kazakhstan

  • Reinhart, C., Rogoff, K. and M. Savastano (2014), “Addicted to Dollars”, Annals of Economics and Finance, 15-1, pp. 1–50

  • Taylor, J. B. (2000), “Low Inflation, Pass-Through, and the Pricing Power of Firms”, European Economic Review, 44, pp. 1389–140

  • Winkelried, D. (2014), “Exchange Rate Pass-Through and Inflation targeting in Peru”, Empirical Economics, 46, pp. 1181–1196

Appendix

Data Sources

Domestic CPI and PPI, as well as the nominal effective exchange rate (NEER) series come from the National Bank of Kazakhstan. For the purposes of this analysis, the inverse of the original NEER series was used, so that a positive change implies depreciation. The source of data on import prices and the industrial production index is the Committee of Statistics. Given the seasonality of industrial output, in most specifications we use seasonally adjusted series. The index of external prices is constructed as a weighted average of producer/wholesale price indices of Kazakhstan’s main trading partners, applying the same weights as in the NEER calculations.1 The primary source for individual price indices is IFS, with Haver Analytics used to gap-fill missing observations.

Robustness Checks

Figure A1.
Figure A1.

Impulse Reponses in a Model with Output Gap

Citation: IMF Staff Country Reports 2017, 109; 10.5089/9781475598759.002.A003

Source: IMF staff calculations
Figure A2.
Figure A2.

Impulse Reponses in a Model with USD

Citation: IMF Staff Country Reports 2017, 109; 10.5089/9781475598759.002.A003

Source: IMF staff calculations
1

Prepared by Rossen Rozenov.

2

See National Bank of Kazakhstan Press Release No. 38 of August 21st, 2015.

3

A substantial part of the CPI basket comprises services that are not related to imports and the share of these services typically increases as per capita income grows. This is one of the factors behind the lower pass-through in advanced economies.

4

The rationale for including the real exchange rate gaps is that in developing countries real exchange rate tend to appreciate as these economies grow faster.

5

Including more lags of the log NEER does not change estimates significantly since estimated coefficients are very close to zero, with alternating signs.

6

Specifically, the vector comprises (i) foreign inflation; (ii) a measure of economic activity; (iii) exchange rate; (iv) import prices; (v) producer prices and (vi) consumer prices.

7

According to Berger and others (2009), in US distribution margins are in the order of 50-70 percent.

8

For both models AIC suggested three lags and tests did not indicate the presence of residual autocorrelation.

1

The weights are available on the National Bank of Kazakhstan website.

Same Series

Other IMF Content

Other Publishers

Cover IMF Staff Country Reports
Republic of Kazakhstan: Selected Issues
Author: International Monetary Fund. Middle East and Central Asia Dept.
Volume/Issue:
Volume 2017: Issue 109
Publisher:
International Monetary Fund
ISBN:
9781475598759
ISSN:
1934-7685
Pages:
61
DOI:
https://doi.org/10.5089/9781475598759.002

Exchange Rate and Various Price Indices (logarithmic scale)

Kazakhstan: Correlations between Inflation and Exchange Rate Changes

Kazakhstan: Impulse Response of CPI to One SD of NEER

Impulse Response of Import Prices to One SD of NEER

Impulse Reponses in a Model with Output Gap

Impulse Reponses in a Model with USD