Working Paper No2009/12
April 2009
Money and exchange rate channels for food and non food inflation: a cointegrated VAR for Zambia
Elva Bova*
Abstract
This paper uses a cointegrated VAR to detect how sensitive Zambian food and non food inflation is to changes in the money supply and in the exchange rate. It finds that the monetary transmission mechanism is weak and effective only for non-food prices, while the exchange rate channel is stronger, especially for food prices. Estimates also indicate that the exchange rate is very sensitive to changes in copper prices. The study suggests, then, that foreign exchange intervention to avoid real appreciations and safeguard competitiveness may not be too inflationary. This is also in the case when the money supply is not sterilised.
Key Words
Exchange rate, Inflation, Monetary policy, cointegrated VAR
* Elva Bova is a Doctoral student of the NCCR individual project on “Primary Commodities” and a PhD candidate in the Economics Department of the School of Oriental and African Studies, University of London.
NCCR TRADE Working Papersare preliminary documents posted on the NCCR Trade Regulation website (< and widely circulated to stimulate discussion and critical comment. These papers have not been formally edited. Citations should refer to a “NCCR Trade Working Paper”, with appropriate reference made to the author(s).
1. Introduction
Due to the recent copper boom the Zambian economy has been exposed to a sudden and significant increase in foreign exchange. This has accrued mainly to the mining companies and large part has been spent into the copper industry for mines expansion and investments. The spending effect associated with the boom has determined a nominal appreciation of the Zambian Kwacha, exacerbated by inflows of foreign capital in the form of aid and of portfolio flows. Moreover, the concession of debt relief in 2005 has released additional foreign exchange in to the economy. As illustrated in different studies (Cali and te Welde, 2007; Fynn, and Haggblade, 2006; Weeks et al 2007; Weeks, 2008; Exports Board of Zambia, 2007) the nominal appreciation in 2005 has dramatically damaged some non traditional exports, such as tobacco, cotton and horticultural products, production of which fell sharply in 2006.
While exporters demanded a more managed exchange rate, the response of the central bank has been to let the currency float, as prescribed by their inflation-focused monetary framework. This, framework maintains price stability as the main objective, to be achieved through a monetary target, and it calls for flexibility in the exchange rate with very limited foreign exchange interventions.
With this focus, foreign exchange intervention to avoid a nominal appreciation during a boom should be avoided, since, when sterilisation of the money supply is not possible, the accumulation of international reserves may be inflationary. Also, with a focus on external competitiveness, an increase in inflation resulting from intervention may, in turn, cause a real appreciation of the currency, simply because domestic goods will become more expensive in the international market. Thus, even under a more managed exchange rate, a boom may damage a country’s competitiveness, through an increase in inflation. Therefore the correct policy would be to let the currency float, since floating would at least avoid an increase in inflation.
In this study we examine whether, in Zambia, a real appreciation would have similarly followed under a more managed exchange rate regime. In this vein, we test for the effectiveness of the monetary transmission mechanism and the exchange rate channel, since these are the mechanisms through which accumulation of foreign exchange may increase domestic prices. To test for effectiveness, we use a cointegrated VAR to examine how sensitive Zambian inflation is to changes in the money supply and in the exchange rate. Since almost 60 per cent of the Zambian CPI is made of food prices, we distinguish between effects on food and on non food prices. The model also includes the copper price and oil price, as exogenous variables, to investigate, firstly, to what extent the Zambian Kwacha responds to changes in the copper price, and, secondly, how much the exchange rate pass through is connected to the oil price.
The paper is organised as follows. The next section presents the theoretical background, while section III offers a brief review of the empirical literature. The unrestricted VAR is described in section IV, while section V illustrates the long and short run restrictions and examines the results. Section VI offers a conclusion and some policy implications.
2. Theoretical background
As postulated by the “Dutch disease” literature, a commodity boom may cause a real exchange rate appreciation with significant damage for the non traditional exports. In the traditional model this impact emerges out of two main effects: a spending effect and a resource switching effect (Corden, 1984; Corden and Neary, 1982; Neary and Van Wijnbergen, 1984). In both cases real exchange rate appreciation is determined by changes in relative prices, that is a relative increase in the price of non tradable goods with respect to the price of tradable goods[1]. This is consistent with a definition of the real exchange rate as the ratio of prices of non tradable goods over prices of tradable goods, which can be considered as a proxy for competitiveness in a small open economy. This first specification of the Dutch disease effects[2]tends, however, to ignore the implications a particular exchange rate regime has on the effects of a boom. This is because one of the main assumptions of the model is of an economy with no money.
In this study we rely, then, on a rather marginal specification of the Dutch disease dynamics, put forward by Edwards in 1989 to examine the impact of the coffee boom in Columbia. This alternative framework distinguishes the impacts a boom may have on a country’s competitiveness in the light of the exchange rate regime.
Under a flexible regime, in fact, an increase in foreign exchange will appreciate the currency in nominal terms, which will be passed on to a real appreciation. Exports will become more expensive, which will damage the country’s external position. On the other hand, under a fixed exchange rate regime, the central bank will need to accumulate foreign exchange in its international reserves so as to keep the parity. To do so it will need to increase the money supply. However, in many developing countries with shallow financial markets[3], the increase in money supply cannot be sterilised and may become inflationary. Inflation, in turn, will appreciate the real exchange rate, since the price of local goods will be higher in the international market[4]. Hence, under either a flexible or a fixed exchange rate a surge in the foreign exchange will appreciate the real exchange rate, which implies that neither arrangement is better than the other; however, a float would at least avoid an increase in inflation.
Whenever the monetary transmission mechanism is weak, however, then there is reason to believe that the real appreciation and consequent loss in competitiveness will be more likely under a float. This may be of relevance in developing countries where there are several reasons why an increase in the money supply may not result in an increase in domestic prices.
Firstly, where the economy is largely a cash economy, any increase in the money supply within the interbank market may be rather marginal compared with the amount of cash circulating into the economy. If this is the case, then, an increase in the money supply may not spur aggregate demand (Saxegaard, 2006). Secondly, if an increase in the money supply does raise aggregate demand, then it may be that this increase does not affect domestic prices. This may be for two main reasons. One is the large food component typical of developing countries’ CPI. Food prices are, in fact, income inelastic, for an increase in demand would be redirected to other goods; therefore a surge in the money supply will not raise the domestic CPI. The other is the fact that the economy does not operate at full capacity. Therefore, an increase in aggregate demand may effectively spur output and prices will remain stable. However, if absorptive capacity is limited in the economy, then prices may increase after all.
However, despite the weak monetary transmission mechanism, foreign exchange interventions may still impact on inflation through the exchange rate channel. This is because in developing countries a large share of goods are imported and their prices are sensitive to changes in the nominal exchange rate. When a central bank conducts foreign exchange interventions to offset a nominal appreciation, it basically cancels out the impact the nominal appreciation would have had on imported prices, namely a reduction of these prices. Thus, with foreign exchange intervention inflation may actually increase, or at least it may not decrease.
In the case of the Zambian economy, it may be that for one or more of the reasons mentioned above, the monetary transmission mechanism is weak. If this is the case, then a more managed float would probably perform better than a float, in as much as it would better safeguard the non traditional exports during the boom.
For this analysis we consider that foreign exchange intervention may impact on inflation through the release of unsterilised money supply (monetary transmission mechanism) and through an increase in the CPI imported component (exchange rate pass-through). As a matter of fact, prices of imported goods will not increase but their level would be higher then the one during a nominal appreciation. Thus we examine whether and to what extent prices of food and non food are sensitive to changes in the money supply and then we consider the extent of the exchange rate pass-through.
3. Empirical literature
To understand how sensitive domestic prices are to changes in the exchange rate and in the money supply we draw on the existing literature on the monetary transmission mechanism in developing countries. The issue of the efficacy of the monetary transmission mechanism has recently been studied with the purpose of detecting whether the link between the money supply and inflation is strong in developing countries. The literature provides mixed evidence with respect to the monetary transmission mechanism in African countries. A seminal study by Canetti and Greene (1992) did not find a common pattern in African countries as far as the causes of inflation are concerned; in some cases the exchange rate channel seems to be more effective, in others the money supply growth has the longer term impact.
As far as Zambia is concerned, there have been different studies on the effectiveness of the monetary transmission mechanism, most of them focusing on the consequences of the exchange rate liberalisation of the early 1990s. A study by Mwenda in 1993 looked at the impact on the effectiveness of monetary policy of switching to indirect monetary policy instruments, with a special focus on growth and variability in broad money and in inflation. The study finds out that the move to indirect instruments for policy has indeed reduced the variability in broad money and inflation. A further study by Adam (1999) found that the variance of currency demand has increased in Zambia since the end of the 1980s. It also observed that stabilisation policy based on controlling reserve money is likely to have an imprecise link to inflation in the short and medium term, which is shown by the fact that short-run forecast variance around the money demand is relatively high.
A more comprehensive study, by Simatele (2004), examined how the effectiveness of the monetary transmission mechanism to the macro-economy has changed with the liberalisation reform, using two different models for the period prior and the period after the boom. The analysis adopted a VAR using the Choleski decomposition to impose restrictions, thus, relying on the assumption that policy does not respond contemporaneously to macro-shocks and that this may be due to information lags. Through impulse responses and variance decompositions the study illustrates that the potency of monetary policy has increased with the reforms, since prices are more responsive to monetary policy shocks. The study also illustrates that the exchange rate seems to be an important variable in the explanation of prices in Zambia.
Probably the most relevant study on the issue, at least for our analysis, is the one by Mutoti (2006). The author tries to investigate short and long term dynamics and transmission mechanisms of post-liberalised Zambia. He does this through a cointegrated VAR, in which restrictions are imposed according to a priori information on the relationships between the variables (Christiano et al., 1994; Leeper et al., 1996). The model is framed in an IS-LM-AS theoretical structure and considers the domestic and foreign (South African) interest rate, money supply (as broad money), output, domestic and foreign prices and the nominal exchange rate Kwacha to the South African Rand. Running the model for the years 1992-2003 Mutoti found that there is a stable money demand relationship, implying that money growth has a predictable impact on the economic activity and also that money demand is sensitive to the interest rate; inflation appears to be associated with excess demand and disequilibrium in the exchange rate. From the impulse responses to a money supply shocks it appears that domestic prices react strongly only in the first period, suggesting that the link between money supply and inflation may be weak. As expressed in Mutoti (2006:18) “(this may give) rise to situations where getting the monetary target does not produce the desired inflation outcome and where money fails to produce reliable signals of the stance of monetary policy. Since food price has the largest share in CPI and the dominant role of exchange rate in inflation dynamics established, sustaining lower inflation in Zambia requires policies meant at boosting domestic food supply and stabilising exchange rate”.
Further to this conclusion, this study analyses the transmission mechanism on the price of food and non food. Similarly to Mutoti, we use a cointegrated VAR, yet we consider money supply and exchange rate as the policy variables and include copper and oil prices.
4. The unrestricted VAR
To formulate a cointegrated VAR for Zambia we consider the following unrestricted structure:
xt = П1xt-1 + П2xt-2 + Θyt-1 + ΦDt + εt;(t=1…..T)(1)
εt ~ Np(0, Σ), (2)
Where xt is a k x 1 vector which includes the following endogenous variables:
M3 = log M3sea – log CPIz, (3)
ER = log CPIz – log CPIus –NEus-z(4)
∆CPI fz,(5)
∆CPI nfz(6)
Where M3is defined as the logarithm of the seasonally adjusted real broad money (M3), deflated by the Zambian headline CPI. ER is the log of the real bilateral exchange rate ZMK-US Dollar[5]. ∆CPI fz,and ∆CPI nfz are, respectively, the rate of change in the food and non food Zambian CPI. For the exogenous variables, yt-1, we use the log of real copper prices, obtained as the nominal market value price of copper in US$ per metric ton divided by the Zambian CPI and the logarithm of the real price of oil, obtained as the oil price in US$ per barrel divided by the Zambian CPI. The data on the Zambian economy are from Bank of Zambia, while the data on copper and oil prices are from the International Financial Statistics database of the IMF.
The model uses a sample of monthly data from April 1996 to April 2008, which corresponds to a period of macroeconomic adjustment subsequent to the adoption of several liberalisation reforms. It also covers the period of the copper boom which starts in approximately 2004. A deterministic trend has been added to the model, which proves to be significant for food and non food inflation and we also account for a break in the trend in December 2005, corresponding to a change in the rate of growth of the money supply. From tests for lag determination reported in annex A the VAR is significant for two lags, which is confirmed by the fact that the autocorrelation in the residuals disappears when accepting two lags(annex A).[6]To assess whether the VAR assumptions are accepted we first conduct a visual inspection of the data and their time-series properties.
Figure 1: The endogenous variables
Source: Bank of Zambia data
The graphs in figure 1 suggest that the series of the endogenous variables display a non stationary behaviour, which may indicate the presence of a unit root in all of them. The non stationarity of the series is however tested within the VAR structure, through the rank test and the standard Dickey Fuller test for unit root is not applied, since its validity is questioned in a multivariate context (Juselius, 2006).
To account for the presence of outliers in the residuals of some of the variables the unrestricted VAR is corrected for blip transitory dummies associated to shocks in non food inflation, and for permanent dummies which are related to money and exchange rate’s behaviour[7]. As discussed in Juselius (2006), the presence of dummy variables in a VAR has a different meaning and implication than in the case of univariate time series analysis. This is because, while the dummy variables eliminate the outlier from a single variable they do not eliminate the impact this outlier has on the relationships among and between variables.
From the residuals analysis in table 1 the model is accepted and correctly specified. According to the LM test for the model with two lags, there is no autocorrelation in the residuals with a rather high p-value of 0.295, although the Ljung Box test does not reject autocorrelation. Normality is not rejected with a p-value of 0.06, and a much higher p-value for the individual series, except for non food inflation where normality is not rejected at 5 per cent confidence level. ARCH residuals are not rejected for multivariate test, but they are rejected with high p-valuesfor the individual series, except for the exchange rate where the p-value is 0.077. Kurtosis and skeweness estimates do not deviate too much from their normal values, of 3 and zero.