Information about Asia and the Pacific Asia y el Pacífico

Chapter 7. Pro-Poor Exchange Rate Policy

Sumio Ishikawa, Sibel Beadle, Damien Eastman, Srobona Mitra, Alejandro Lopez Mejia, Wafa Abdelati, Koji Nakamura, Il Lee, Sònia Muñoz, Robert Hagemann, David Coe, and Nadia Rendak
Published Date:
February 2006
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details
II Houng Lee and Srobona Mitra 

This chapter reviews exchange rate policy options that could minimize any adverse impact on the poor from an exogenous shock such as the elimination of the quota system in early 2005. As long as the buildup of pressures on the balance of payments is modest relative to the level of official reserves, maintaining a stable exchange rate could reduce potential adverse effects on incomes of the poor. In Section A, we present stylized facts, unique to Cambodia. Section B presents results of estimation of exchange rate pass-through to better assess policy implications. In Section C, we provide pro-poor exchange rate policy options.

A. Stylized Facts

Cambodia is a de facto dollar economy. Most prices, except for some prices of nontradable goods and services in rural areas, are denominated in U.S. dollars, and up to 95 percent of total liquidity (including estimated cash in circulation) is in U.S. dollars. The domestic currency, the Cambodian riel (CR), is used mainly by the rural population as a medium of exchange, by the urban population as “coins” complementing U.S. dollars in circulation, and by the government, which spends more in riel than it collects.

As such, it is difficult to interpret the exchange rate as the relative value of currencies of two countries that would normally change in response to differentials in inflation and productivity growth, or changes in the terms of trade. But rather, the exchange rate reflects changes in demand for riel by a small fraction of the urban population who hold riel cash balances for transaction purposes. Although most of the poor in the rural areas hold riel cash balances, they appear not to contribute much to exchange rate changes due to limited information and access to the foreign exchange market. The demand for riel is normally met by government spending through its extensive network of treasury branches across the country. In 2002, for example, the government injected about CR 0.3 trillion (i.e., spending in riel net of collection in riel), of which about two-thirds was retained as cash in circulation, matching the trend increase in riel demand in tandem with economic growth. The remaining amount was converted back to U.S. dollars by the private sector (Figure 7.1). Given the relative stability of growth of riel demand, barring any adverse sentiment, excessive cash injection of the government financed by bank borrowing is usually translated into a depreciation of the exchange rate, as the above chart shows, unless the National Bank of Cambodia (NBC) sells its foreign exchange, which it does not often do.

Figure 7.1.Changes in Net Claims on Government (NCG) and Exchange Rate

Sources: Cambodian authorities; and IMF staff estimates.

The demand for riel is also sensitive to noneconomic news such as political developments. Any negative news that raises country risk will immediately lead to further dollarization and, on a much larger scale, to capital outflows. This was evidenced when the total stock of foreign currency deposits dropped by 20 percent in the course of one to two weeks during the July 2003 elections. Moreover, bank owners effectively withdrew their capital—which could not actually be withdrawn unless they liquidated the bank—in the form of bank loans to themselves. An increase in the currency risk, such as an increase in the exchange rate volatility, will prompt those holding riel to convert into U.S. dollars, as those who have access to the foreign exchange market are more concerned with retaining the value of their wealth in U.S. dollar terms. Only a small amount of excess supply or demand could affect a change in the exchange rate because the market is very shallow. With this in mind, a National Bureau Economic Research working paper reviewed about 85 countries that tried to de-dollarize (Reinhardt, Rogoff, and Savastano, 2003), and found that only two of them managed to reduce the foreign currency deposit ratio significantly and keep it low for some time through reduced exchange rate volatility.

Inflation in trading partners in U.S. dollar terms has moved broadly in tandem with domestic prices (in riel), except for the Asian crisis period.38 During 1997–98, although the weighted average of trading partners’ inflation in their respective currencies rose to 7–8 percent, in U.S. dollar terms, inflation declined to below minus 30 percent due to the large devaluations of the exchange rates against the U.S. dollar (see Figure 7.2a). Once trading partner inflation is converted into riel, the large difference during the Asian crisis is sharply reduced (Figure 7.2b), although the initial sharp depreciation in late 1997 (especially of the Thai baht) and the sharp appreciation in 1998 are clearly evident in the remaining gaps during this period.

Figure 7.2.CPI Inflation and Exchange Rate Depreciation

(In percent)

Sources: Cambodian authorities; and IMF staff estimates.

The overall pass-through from exchange rate changes to domestic inflation in a dollarized economy needs to consider the “accounting” effect of exchange rate movements into domestic price (expressed in riel) in addition to policy and other responses of different components of CPI to exchange rate changes.39 To illustrate this point, denote PT = P* where PT is the price of tradable goods and P* is the world price, both in U.S. dollars.40 The price level in turn is defined in U.S. dollars as P=PTεPN(1ε) where 0 ≤ ε ≤ 1 and PN=PNUSβ(PNCRe)(1β). PNCR is the price of nontradables denominated in riel; β is the share of the nontradable goods denominated in U.S. dollars where 0 ≤ β ≤ 1 and e is the exchange rate where de/dt > 0 implies a depreciation. However, since the official CPI is collected in riel, the observed price level is PCR=ePTεPN(1ε), which encapsulates the “accounting” effect.

Changes in the domestic price level can thus be accounted for by (1) nominal exchange rate changes, (2) changes in PT approximated by partner-country inflation41, (3) changes in prices of nontraded goods indexed and priced in U.S. dollars, and (4) change in the riel-denominated price of nontraded goods:

where X˙=dX/X.

The immediate pass-through from exchange rate changes to domestic inflation in the accounting sense is β + ε – βε. The size of this can vary depending on the extent of currency substitution, indexation of domestic prices to the dollar, the size of the tradable goods sector, etc. As examples:

  • If all prices are indexed and denominated in U.S. dollars, β = 1 pass-through is full (=1).
  • If all nontraded goods are denominated and priced in riel, β = 0 immediate pass-through is ε.
  • If the tradables sector is very large or, as a simplification, if there were no nontraded goods, ε = 1 (and β = 0) : pass-through is full.
  • If the tradables sector is very small or, as a special case, nonexistent, ε = 0 : pass-through is β. If, in addition to a small tradables sector, most prices of nontradables are denominated in riel, then pass-through is 0.

The observed pass-through could be different from this “accounting” effect owing to movements of prices arising from policy responses of the central bank to exchange rate movements, unrelated but concomitant domestic demand and supply shocks, and decisions by producers on whether to increase prices of goods (depending on elasticity of demand) in response to higher cost of imported inputs.42

B. Estimating Pass-Through of the Exchange Rate

A vector auto regression (VAR) model is used to estimate the extent of pass-through from exchange rate depreciation to domestic inflation, using monthly data from January 1996 through December 2003. Owing to data constraints, only a trivariate VAR is considered, recognizing that there could be external factors other than exchange rate that could affect the domestic price level. The U.S. dollar inflation rate of trading partner countries is considered as a suitable candidate that might affect domestic inflation. A real variable, like the output gap, is left out because of the lack of data.43

Granger causality tests show that both exchange rate depreciation ê and partner inflation π* Granger-cause domestic inflation rate π; however, inflation does not Granger-cause ê and π*. The variables ê and π* do not Granger-cause each other.

To retrieve the structural shocks to each variable, short-run restrictions are imposed—shocks to ê and π have no contemporaneous effect on π*, and shocks to π have no contemporaneous effect on ê. These restrictions have the same effect as that of ordering the variables as π*, ê, and π.44 The lag-length of 2 was chosen using the Schwarz information criterion (SIC), although the results are robust to inclusion of higher lags.

Following a 1-standard deviation shock (increase) to the exchange rate (a depreciation), domestic inflation goes up—with a maximum effect at four months. The associated pass-through45 from exchange rate depreciation to domestic inflation is 18 percent on impact, around 41 percent by the end of the first year, and 51 percent at the end of two years following the shock (Figure 7.3). The negative effect of partner country inflation shocks on domestic inflation appears to follow the strong negative relationship—arising from common shocks to regional exchange rates against the U.S. dollar—observed in some periods during the aftermath of the Asian crisis.

Figure 7.3.Pass-Through of Exchange Rate and Partner Inflation

(In percent)

Sources: Cambodian authorities; and IMF staff estimates.

To assess whether the large shocks in the Asian crisis are producing this result, the VAR was rerun using a sub-period of January 1999 to December 2003. The results now show that the impulse response of inflation owing to partner inflation shocks is positive, mostly at five to nine months. However, the impulse response because of exchange rate shocks is now negative, but not significantly different from zero. The result could reflect one or more of the following: (1) greater domestic policy focus on bringing down inflation in Cambodia in the post–Asian crisis period; (2) an increased weight of nontraded goods denominated in local currency (unobserved) in the CPI in recent years; or (3) other large shocks arising out of the nontraded goods sector.

The source of the exchange rate shock is mainly external. This will result in a pure accounting effect in the first round—given nontradable prices are sticky in riel, and tradable goods prices are set by partner countries in U.S. dollars. In U.S. dollar terms, real income of the poor will decline by the exchange rate depreciation effect. But in the subsequent rounds there could be some adjustment according to whether the exchange rate depreciation is perceived to be permanent or temporary. If this depreciation is seen as very short lived, then there should be no adjustment in wages, and hence prices, of the nontradables sector.

C. Pro-Poor Exchange Rate Policy

An exogenous shock such as the elimination of the quota system in 2005 could worsen the trade deficit by about $120 million (equal to about 20 percent of net international reserves). While the impact on banks is likely to be limited, for example, to a slower growth in U.S. dollar deposits,46 confidence could be shaken, leading to further dollarization and capital outflows. The NBC could respond either by allowing the exchange rate to depreciate or by defending the rate, which would not be difficult given the limited amount of foreign exchange reserves such a defense might require. However, the implications of these policy options are more widespread.

Case 1. No foreign exchange market intervention: Suppose the NBC does not intervene and the exchange rate depreciates, say by 10 percent. The net impact on the fiscal position is estimated to be about 0.2 percent of GDP of additional revenue in domestic currency terms (see Table 7.1). The net fiscal gain generated from the depreciation will not change much in the short run even if the government were to simply spend more in riel as, except for civil service wages and social transfers, government spending is largely tied to U.S. dollar prices.

Table 7.1.Budgetary Outlays, 2002(In percent of GDP)
Domestic CurrencyForeign CurrencyTotalImpact of 10% depreciationTotal
Foreign financing0.
Domestic bank financing–1.10.0–1.10.0–1.1
Domestic nonbank0.
Sources: Cambodian authorities; and IMF staff estimates.
Sources: Cambodian authorities; and IMF staff estimates.

Case 1a: Since the fiscal position should not be procyclical, especially at a time when real GDP growth is expected to slow, the government would need to raise civil service wages and social transfers at least to a level that would maintain a balanced recurrent budget. In the first instance, the result of this policy will be as follows: (1) the nongarment formal sector wealth (financial assets) and income will be unchanged, (2) the financial wealth (which consists only of riel cash holding) of the rural poor will be reduced by 10 percent in U.S. dollar terms, and (3) their income in U.S. dollar terms will be adversely affected to the extent that there is less-than-perfect pass-through and some of the nontradable goods and services they provide are denominated in riel terms. To illustrate the point, define real income of the rural poor as qNPN+qTPTP where qNPN is the nominal income from producing nontradable goods and services. Since, as noted earlier, PN is partly denominated in riel terms (assume for the sake of simplicity that all nontradable goods are denominated in riel), real income in U.S. dollar terms is defined as qNPNCR(1e)+qTPTP. To the extent there is less than full pass-through of e to PNCR, the nontradables component of their real income qNPNCR(1e)P will decline as e increases. The reduced consumption of the poor will marginally contribute to further adjustment of the balance of payment.

Case 1b: Alternatively, the government could allow the fiscal position to be procyclical. In this case, civil service wages and social spending will decline, leading to an adjustment through contraction of aggregate demand. In this case, (4) civil servants will be relatively poorer, and (5) the recipients of social spending will be adversely affected.

Case 2: Foreign exchange market intervention: If the NBC intervenes to maintain a stable exchange rate, international reserves will decline. At least in the first instance, however, it will not lead to any of the results noted in (2) through (5).

Under both cases, the secondary impact will be lower GDP growth, hence reduced income, mainly in urban areas. Rural areas will be affected to the extent that there is less transfer from urban workers’ income to their families.

D. Conclusion

While exchange rate flexibility should be maintained in order to absorb exogenous shocks, careful consideration will need to be given to the redistributive effects of the exchange rate policy. This is particularly so in cases where the potential impact of a shock on the exchange rate and on international reserves is not expected to be large, and where an adjustment of the exchange rate will not have significant impact in helping to absorb a change in the structure of the economy, given the high degree of dollarization and its asymmetry across income groups.


The main trading partners, which together account for more than 60 percent of Cambodia’s total imports, are Thailand, Singapore, Hong Kong SAR, Korea, and Vietnam. The monthly trading partners’ inflation data were compiled from monthly CPI and exchange rate data, but using annual trade weights.


These effects would normally depend upon the source of the exchange rate shock.


The real exchange rate r is defined as r = PN/P* where PN is the price of nontraded goods.


This assumes that imported-traded goods are priced in the producers’ currency (here, U.S. dollars).


In general, price and exchange rate movements are likely to be positively correlated when they are caused by common factors, like monetary shocks. If the shock is purely external, like a shock to partner country inflation caused (for example) by partner country monetary policy, the correlation between price and exchange rate responses should be near zero. On the other hand, when the authorities target inflation, and use contractionary monetary policy to reduce upward pressures in prices, the exchange rate would tend to appreciate (resulting in a negative correlation with prices).


Various unit root tests confirmed the presence of unit roots in the riel-U.S. dollar exchange rate and domestic CPI series; therefore, 12-month changes in the logarithm of these variables have been used for the quantitative analysis. The partner country inflation rate series was found to be stationary.


This alternative to the recursive Cholesky orthogonalization requires imposing enough restrictions to identify the orthogonal (structural) components of the error terms. An ordering with ê first and π* second does not change the results in this case.


Pass-through at time t is calculated from the accumulated impulse responses as i=1tIRi(πtoêshocks)i=1tIRi(êtoownshocks); IR is impulse response.


During 2001–02, balance of payments surpluses were partly absorbed by large increases in foreign currency deposits.

    Other Resources Citing This Publication