Abstract
This Selected Issues paper analyzes reserve adequacy in Mexico. Reserve adequacy has been of renewed interest, as the authorities have introduced a new rules-based mechanism of U.S. dollar sales to reduce the rate of reserve accumulation. The paper examines the recent experience with Mexico’s fiscal management tools in light of the need for further consolidation. It reviews evidence on the cyclical behavior of fiscal policy in Mexico, finding that it has generally been procyclical since the early 1990s. The paper also examines the determinants of inflation in Mexico since 1997.
IV. Dollarization in Mexico1
This chapter extends previous work on dollarization in Mexico in three respects. First, it covers the period 1990–2003, which has been characterized by a clear decline in dollarization. Second, we experiment with a broader definition of dollarization than has been used in previous studies. Third, a variable proxying for vulnerability to external shocks is included to assess its significance in explaining the reversal of dollarization using an extended portfolio balance model. We find that reduced vulnerability to external shocks is significant in explaining the reversal. Regarding other explanatory variables, we find that the sign on the expected depreciation of the peso is also significant but negative, which we interpret as reflecting the effect of prudential regulations in limiting dollarization.
A. Overview
84. Despite enhanced financial stability, informal dollarization has remained high or even increased since the early 1990s in several Latin American countries, but has declined in Mexico. In Bolivia and Peru, for instance, dollarization rose during the 1990s and has declined only slightly since 2000, with overall dollarization ratios still high at about 65 percent for Peru and 92 percent for Bolivia. The ratio of foreign currency deposits (FCDs) to total deposits remains even higher, in the 80–90 percent range, for these countries. These experiences contradict the common assumption that the substitution process between currencies is symmetric, responding to changes in the determinants of dollarization— particularly financial instability—in both directions (Buchs, 2000). In this regard, Mexico stands out in that the extent of dollarization has been more closely linked to periods of financial instability, and has reversed as stability has been restored. The experience since 1996–97 is a particularly striking example (Figure 1).2
Mexico: Foreign Currency Deposits to Broad Money (M4)
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; and Fund staff estimates.Mexico: Foreign Currency Deposits to Broad Money (M4)
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; and Fund staff estimates.Mexico: Foreign Currency Deposits to Broad Money (M4)
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; and Fund staff estimates.85. This paper studies the reversal in dollarization since 1996 in Mexico, an episode as yet not covered in the literature. Several studies have examined dollarization in the 1970s and 1980s (see below), but there does not appear to have been any analysis of the reversal since 1996. We fill this gap, and extend previous work in three respects: (i) by analyzing the factors that could explain the reversal since 1996; (ii) by using a broader definition of currency substitution than in previous studies on Mexico includes dollar holdings of Mexicans in U.S. banks; and (iii) by explicitly modeling external vulnerability to examine its significance in explaining the reversal.3
86. The paper is structured as follows. Section B provides a brief review of the literature on currency substitution in Mexico. Section C describes the trends in dollarization from 1977 until the mid-1990s, and the impact of these developments on residents’ decisions to hold foreign currency. Section D focuses on the post 1994–95 crisis period, describing broad macroeconomic developments and their role in explaining the reversal in dollarization. Section E describes the econometric model and estimation results, while Section F provides some general conclusions and policy implications.
B. Literature on Currency Substitution
87. Several explanations have been advanced to explain the persistence of informal dollarization in some countries. Among the reasons cited commonly for this asymmetric response are the following:
Episodes of high inflation tend to remain in memory for longer than periods of low inflation, and are assigned more weight in portfolio allocation decisions (Ramirez-Rojas, 1985; Clements and others, 1992 and Agenor and Khan, 1992).
Adjustment costs create a hysteresis, or “ratchet” effect, so that once domestic residents convert monetary holdings into dollars, it is costly for them to return to the local currency, even after inflation decreases. Thus, temporary increases in inflation may have a permanent effect on the demand for domestic money (Dornbusch and Reynoso, 1989; Guidotti and Rodriguez, 1992; Clements and Schwartz, 1993).
The existence of “network externalities” means that, as more people use dollars, the transactions costs of using dollars decrease (Uribe, 1995).
Episodes of financial instability create permanent reputational differences between “hard” and “soft” currencies. Once agents have shifted to holding hard currencies as a result of high inflation or exchange rate depreciation, they resist moving back to soft currencies even though financial conditions stabilize (Yotopoulos, 1997).
88. In the case of Mexico, some empirical studies have found exchange rate expectations to be a key determinant of currency substitution. Previous empirical studies on dollarization in Mexico include Ortiz (1983), Ramirez-Rojas (1985), Rogers (1992a, b), and Gruben and Welch (1996). Ortiz examined currency substitution during 1933–80, focusing on the behavior of the ratio of foreign currency demand deposits to domestic currency demand deposits. Using a quarterly demand for money function, he found that this ratio was positively related to expectations of exchange rate devaluation and foreign exchange risk. Ramirez-Rojas compared the experience in Argentina, Mexico, and Uruguay from 1970 through the early 1980s. Defining dollarization as the ratio of foreign currency deposits (FCDs) to the total money stock (which included FCDs), he also found that the expected change in the exchange rate (proxied by the inflation differential) increased dollarization.
89. On the other hand, Rogers (1992a, b), using data for 1978–85, found a statistically significant negative relationship between dollarization and the expected rate of peso depreciation. He interpreted this to mean that depositors anticipated a breakdown of full convertibility between peso and dollar deposits as expectations of peso depreciation increased, precipitating a run on foreign currency deposits. Gruben and Welch (1986) examined the effect of deteriorating bank loan quality on the dollarization ratio, finding a significant negative relationship between dollarization and nonperforming loans, and the conventional positive relationship between expected peso devaluation and dollarization. They concluded that the results in Roger’s study represented a case of a “missing variable”— specifically bank asset quality.4
C. Definition and Trends in Dollarization in Mexico
Definition of dollarization
90. Most studies have measured dollarization as the ratio of foreign currency deposits (FCDs) in domestic banks to broad money (M2). An important issue is whether this measure may underestimate the full extent of dollarization given the existence of crossborder dollar deposits held by Mexican residents, particularly in the United States. Such holdings could reflect the desire of Mexican residents to insure against the risks of confiscation by domestic authorities, the inability to hold FCDs in the domestic banking system, or simply increased integration with the United States following entry into NAFTA in 1994.
91. In this paper, we explicitly take account of cross-border deposits. We define foreign currency deposits of Mexicans as the sum of FCDs in the domestic banking system plus cross-border deposits held by Mexican residents in U.S. banks. In the same vein, we use a wider definition of broad money, defined as broad money (M4) plus the cross border deposits of Mexicans in U.S. banks.5 This provides a more comprehensive measure of dollarization, including both portfolio and transactions motives for holding foreign currency. As seen in Figure 2, the inclusion of cross-border deposits significantly affects the level of the dollarization ratio, although the broad trends are the same, with the exception of the period during the 1994–95 crisis. This divergence is discussed below.
Mexico: Dollarization Ratios, 1985-2003
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; U.S. Department of Treasury; and Fund staff estimates.Mexico: Dollarization Ratios, 1985-2003
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; U.S. Department of Treasury; and Fund staff estimates.Mexico: Dollarization Ratios, 1985-2003
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; U.S. Department of Treasury; and Fund staff estimates.Trends in dollarization
92. The economic crisis of 1976 led to an extended period of macroeconomic instability. In the early 1970s, Mexico’s inflation differential with the United States was low and stable, oscillating around 2 percent; the exchange rate was fixed; there were no capital controls; and dollarization was almost absent. The economic crisis of 1976, however, led to the first devaluation of the peso since the 1950s, and the fixed exchange rate system was replaced with a flexible regime. Mexico started an adjustment program, supported by the Fund, aiming at improving the fiscal situation, but the discovery of large petroleum reserves in 1977 led the government to embark on an ambitious public sector-led growth strategy. Although this strategy was initially successful in generating growth, which averaged 8 percent during 1978–81, the public sector borrowing requirement doubled from an average of 7 percent of GDP in 1976–78 to 14 percent of GDP in 1981, despite record oil receipts. This was accompanied by an appreciation of the real effective exchange rate, accelerating inflation and a deterioration in the current account.6 While the public sector deficit was initially covered by credit from the central bank, by 1981 external loans had become the main source of financing.
93. The inconsistency of the exchange rate policy with the expansionary fiscal stance led to an increasing realization that the situation was untenable, and in turn increased dollarization. The premium on the forward exchange rate started to rise and residents switched from domestic to foreign currency, with dollarization reaching 28 percent of M4 in 1981. A balance of payments crisis ensued in 1982, with two sharp devaluations in February and in August. The first devaluation temporarily reduced speculative pressures on the peso, but as capital flight intensified and international perceptions of Mexico’s ability to repay its external debt worsened, access to capital markets was severely reduced. This forced the authorities to devalue again in August 1982.
94. At the same time, FCDs were declared inconvertible into pesos at the market rate. The government announced a forced conversion of the outstanding stock of FCDs into domestic currency at a pre-determined rate, and imposed foreign exchange and capital controls in an attempt to resolve the debt crisis amidst dwindling reserves.7 The observed dollarization ratio dropped from nearly 30 percent in August 1982 to 3 percent by December 1982, despite high inflation and a negative interest rate differential with the United States. The remaining FCDs in the banking system reflected mainly official holdings.
95. In December 1982, Mexico embarked on another stabilization program supported by the IMF, but it was derailed by a series of external shocks. The main objective of this program was to reduce inflation, and improve the balance of payments situation through cuts in public sector deficits and reduced monetary expansion. The primary deficit started to improve from 1983 onwards, and recourse to central bank financing was severely curtailed. However, a number of exogenous shocks led to a weakening of the fiscal position and inflation began to rise. Meanwhile, restrictions on FCDs were eased in 1985 and eliminated in 1987, establishing full convertibility once again. As a result, FCDs started to grow.
96. In December 1987, the government adopted a radically altered stabilization strategy—the Pacto —and in March 1988 introduced a fixed exchange rate system in a bid to stabilize the economy. The peso was fixed to the U.S. dollar from March to December 1988. Nevertheless, dollarization continued to increase. This may have been due to what Khor and Rojas-Suarez (1991) have called the “peso problem,” that is, the fixed exchange rate system did not instill confidence, and expectations of the future spot rate consistently over-estimated the actual future spot rate as market participants believed that the fixed exchange rate system would not last for long. As a result, people saw continued merit in maintaining high levels of foreign currency deposits. Under the circumstances, given the continued divergence of market expectations of the exchange rate from the announced fixed rate, the authorities allowed the exchange rate to depreciate at a pre-determined rate and fiscal restraint was intensified. The general policy strategy was successful in arresting inflation, which dropped from 160 percent in 1989 to 12 percent in 1992, and further to single digits in 1993. The dollarization ratio, which had jumped to about 15 percent in 1989, declined to under 10 percent. Mexico was seen as entering a period of sustained prosperity.
97. The next crisis came in 1994, however, when Mexico experienced a series of financial shocks, reflecting in part adverse political events, an over-appreciated exchange rate, and loose fiscal policy.8 There was also increased competition for funds from other emerging market economies, combined with higher U.S. dollar interest rates in the world market. In late 1994, the peso came under speculative attack. In the face of continued market pressure, the authorities were forced to devalue the peso by raising the upper limit of the exchange rate band to 15 percent. In the event, this failed to stem the outflow of reserves, which dropped to about US$6½ billion at end-December 1994 compared with about US$17 billion just three months previously. On December 22, 1994 the peso was allowed to float. Nonetheless, it continued to weaken, as people realized that the government lacked the resources to pay US$29 billion in Tesobonos falling due in 1995. This uncertainty was quickly reflected in a rising dollarization ratio, which jumped to almost 30 percent in 1994–95. In an effort to regain macroeconomic stability, the government agreed on an emergency economic plan with labor and business comprising tight monetary and fiscal policy, accompanied by major structural reforms, including privatization of banks, increases in government controlled prices and a higher VAT rate.
98. The adoption of tight monetary and fiscal policies resulted in a decline in inflation, while improved public finances and the floating rate system helped to restore confidence in macroeconomic policies. In 1996, the Bank of Mexico (BOM) revised prudential regulations on banks with respect to FCDs, which had the effect of discouraging such deposits (see below). As a result, Mexico has witnessed a gradual and sustained decline in the FCD ratio since 1997. The decline in the dollarization ratio, despite large capital inflows, probably reflects a combination of prudent fiscal and monetary policies and combined with prudential regulations on the banking system, which promoted macroeconomic stability and improved confidence.
99. Foreign currency loans in Mexico have also declined significantly since the 1994 crisis. The ratio of foreign currency loans to total resources—include resources available in the domestic financial system (M4 plus bank’s external debt)—dropped from nearly 60 percent in 1994 to under 10 percent in 2003 (Figure 3).9 Figure 3 also suggests that foreign checking accounts in domestic banks closely mirror developments in foreign trade.
Mexico: Credit in Foreign Currency, 1990-2003
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; staff estimates.Mexico: Credit in Foreign Currency, 1990-2003
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; staff estimates.Mexico: Credit in Foreign Currency, 1990-2003
Citation: IMF Staff Country Reports 2004, 250; 10.5089/9781451955613.002.A004
Source: Bank of Mexico; staff estimates.D. What Explains the Reversal in Dollarization in Mexico?
100. Two important developments since the 1994–95 crisis may help explain the reversal in dollarization in Mexico: (i) the introduction of strict prudential regulations affecting FCDs, beginning in 1996; and (ii) orthodox economic policies that resulted in declining inflation, a generally positive real interest rate differential between Mexico and the United States, and increased confidence in macroeconomic policies. The latter is also reflected in progressively improving credit ratings since the adoption of the floating exchange rate regime in late 1994.
Regulations
101. Under current regulations instituted in 1996, banks are permitted to offer foreign currency deposits in checking accounts payable in Mexico provided the holders of such accounts are either: (i) residents of the twenty-kilometer northern border region of Mexico, or those living in Baja California or Baja California Sur; or (ii) firms domiciled in Mexico. Commercial banks are also permitted to offer foreign currency deposits for foreign firms with operations in Mexico provided such deposits are payable abroad. In addition, commercial banks can offer foreign currency accounts in other countries for residents of Mexico. However, there are prudential regulations to ensure that banks invest FCDs in liquid assets, and to prevent maturity mismatches between foreign currency assets and liabilities. The regulations imply that banks can only earn marginal returns on such deposits, reducing their incentives to mobilize them. The more important regulations are described in Box 1, while Box 2 compares these with regulations in Bolivia and Peru, which are still characterized by high dollarization.
Confidence in macroeconomic policies and reforms
102. Economic stability is likely to be a necessary, if not sufficient, condition for a reversal in informal dollarization. The focus of macroeconomic management in Mexico since the 1994–95 crisis has been to ensure low inflation and stable exchange rates, limit fiscal deficits, manage debt prudently to reduce vulnerabilities, and implement structural reforms and integration with NAFTA. One of the most important steps undertaken after the collapse of the peso in late 1994 was to spell out clearly that the principal focus of monetary policy was on attaining medium-term price stability, and that banking sector problems would be addressed by specific programs, with the cost to be assumed by the fiscal authority.
Mexico: Prudential Regulations on Banks
The following prudential regulations are currently in effect on foreign currency assets and liabilities:
For all foreign currency liabilities with a maturity of up to 60 days, banks must maintain liquid assets equivalent to 50 percent of liabilities with a maturity of 1 day, 48.3 percent of liabilities with a maturity of 2 days, and so on. Liquid assets are defined as (i) cash in US$ or in any other freely convertible foreign currency, (ii) T-bills, T-notes and T-bonds issued by the U.S. government with maximum maturity of one year, and (iii) one-day notice deposits in foreign financial institutions rated P-1 by Moody’s Investors Service or A-1 by S&P. These liquid assets cannot be used as collateral, credits, repurchase agreements, or any similar operations that restricts their availability.
The remaining foreign currency holdings can be used for granting credits or investing in other assets based on the guidelines in the Law on Credit Institutions and other applicable provisions. However, there are limits on the amount that banks are allowed to lend to individual borrowers.
Irrespective of the counterparty’s residence, total liabilities of commercial banks denominated in or referred to in foreign currency (excluding cash and highly liquid assets as determined by the BOM) must not exceed an amount equal to 183 percent of the capital base (in line with the Basel Accord) of the respective bank.
At the close of daily operations, assets in foreign currency must be equal to or greater than liabilities in foreign currency in each of the four groups, where assets and liabilities have to be classified into the following four groups—with maturity of 1 day; 8 days or less; 30 days or less and 60 days or less.
Net open foreign exchange position of a bank must not exceed 15 percent of its capital.
Credit institutions have to report daily (no later than 9:00 a.m. the following day) on all their foreign currency transactions.
Mexico: Current Prudential Regulations on Foreign Currency Holdings by Banks Selected Latin American Countries
| Type of restriction/regulation | Mexico | Bolivia | Peru |
|---|---|---|---|
| Foreign exchange accounts permitted | (i) Residents in border areas (ii) Embassies, consulates and international organizations (iii) Firms resident in Mexico (iv) Other firms but as long as payments are made abroad | Residents and nonresidents | Residents and nonresidents |
| Accounts in domestic currency convertible to foreign currency | Not permitted. | Permitted (for both residents and nonresidents). | Permitted (for both residents and nonresidents). |
| Lending in foreign currency | Total liabilities of commercial banks denominated in or referred to in foreign currency (excluding cash and highly liquid assets as determined by the BOM) must not exceed an amount equal to 183 percent of the basic capital stock (in line with the Basel Accord) of the respective bank. | Permitted. Since (October 2002), banks lending criteria must consider financial risks associated to currency mismatch at the firm level. | |
| Liquidity requirements on FCDs | For all foreign currency liabilities of up to 60 days banks must maintain liquid assets (see Box 1) | 10 percent on time deposits with a maturity of up to 720 days, none beyond that. | 20 percent of short-term foreign currency liabilities must be held in liquid assets. |
| Reserve requirements | No reserve requirements. | 2 percent reserve requirement on FCDs with a maturity of up to 360 days. | FCDs are subject to a marginal reserve requirement of 20 percent. Currently, the average rate is about 30 percent. |
| Open foreign exchange positions | Limit is 15 percent of net capital. | Limit is 80 percent of the value of banks’ net worth minus their fixed assets for excess buy position, and 20 percent for excess sell position. | Prudential limit of 100 percent of the net worth over the long foreign exchange position and a limit of 2.5 percent over the short foreign exchange position of the financial institution. |
| Other restrictions | Assets must be equal to or exceed liabilities in each of the four maturity groups. | None | None. |
| Reporting requirements | All foreign exchange transactions to be reported daily. | None. | All foreign exchange transactions to be reported daily on an aggregated basis. |
| Type of restriction/regulation | Mexico | Bolivia | Peru |
|---|---|---|---|
| Foreign exchange accounts permitted | (i) Residents in border areas (ii) Embassies, consulates and international organizations (iii) Firms resident in Mexico (iv) Other firms but as long as payments are made abroad | Residents and nonresidents | Residents and nonresidents |
| Accounts in domestic currency convertible to foreign currency | Not permitted. | Permitted (for both residents and nonresidents). | Permitted (for both residents and nonresidents). |
| Lending in foreign currency | Total liabilities of commercial banks denominated in or referred to in foreign currency (excluding cash and highly liquid assets as determined by the BOM) must not exceed an amount equal to 183 percent of the basic capital stock (in line with the Basel Accord) of the respective bank. | Permitted. Since (October 2002), banks lending criteria must consider financial risks associated to currency mismatch at the firm level. | |
| Liquidity requirements on FCDs | For all foreign currency liabilities of up to 60 days banks must maintain liquid assets (see Box 1) | 10 percent on time deposits with a maturity of up to 720 days, none beyond that. | 20 percent of short-term foreign currency liabilities must be held in liquid assets. |
| Reserve requirements | No reserve requirements. | 2 percent reserve requirement on FCDs with a maturity of up to 360 days. | FCDs are subject to a marginal reserve requirement of 20 percent. Currently, the average rate is about 30 percent. |
| Open foreign exchange positions | Limit is 15 percent of net capital. | Limit is 80 percent of the value of banks’ net worth minus their fixed assets for excess buy position, and 20 percent for excess sell position. | Prudential limit of 100 percent of the net worth over the long foreign exchange position and a limit of 2.5 percent over the short foreign exchange position of the financial institution. |
| Other restrictions | Assets must be equal to or exceed liabilities in each of the four maturity groups. | None | None. |
| Reporting requirements | All foreign exchange transactions to be reported daily. | None. | All foreign exchange transactions to be reported daily on an aggregated basis. |
103. Since 1996, the Bank of Mexico has pursued a gradual process of disinflation. Monetary policy since then has converged on a framework that includes two main elements—an annual inflation target, and the use of a daily objective for settlement balances, the “corto”, to affect interest rates in the pursuit of inflation targets—within a floating exchange rate regime. Importantly, the inflation objective, has now become the nominal anchor for the economy.
104. Currency instability is one of the principal factors causing individuals to seek a “safe haven” in a foreign currency. As noted above, the peso was allowed to float on December 22, 1994 following a failed effort to defend the currency by widening the exchange rate band. This ended the era of speculative attacks on the peso and, combined with relatively low inflation, the exchange rate has been much more stable. For example, the exchange rate depreciated by a mere 3 percent in 1996–97 compared with 44 percent in 1995. In 1998 the exchange rate depreciated by about 22 percent, mainly reflecting the external environment associated with the Russian and Brazilian crises. Since then the exchange rate has moved in a narrow range and even the recent Argentine crisis had little effect. This has played a strong role in returning confidence in the domestic currency, and may have contributed to the gradual de-dollarization of the economy.
105. The banking sector suffered a sharp contraction during the 1994–95 crisis, in response to which the authorities took a number of steps to limit the impact of the crisis. In addition to taking over some banks and liquidating others, the authorities designed debtor support programs, mainly involving incentives for debt restructuring; lengthening loan maturities and assuring a constant real interest rate on debt through an indexation mechanism based on CPI inflation. Although the fiscal costs of rescuing the banking sector were significant, this approach moderated the impact of the crisis on the domestic economy, and banks have strengthened their balance sheets considerably.10
106. International financial markets have also responded favorably to the improved performance of the Mexican economy, as evidenced by stronger ratings by credit rating agencies. Most recently, Mexico’s foreign-currency denominated bonds were upgraded to investment-grade status by both S&P and Fitch IBCA, and Moody’s upgraded Mexico’s sovereign debt to investment grade in February 2002. The recent prepayments of Brady bonds resulted in a further improvement in Mexico’s sovereign rating. This has widened the investor base, and interest rate differentials between Mexico and the United States have declined further. International bond spreads, as represented by EMBI+ for Mexico, have also been generally declining, although they increased temporarily during the Russian and Brazilian crises. The EMBI+ spread for Mexico, as of August 2003, stood at just over 200 basis points, compared with an average of about 500 basis points for all emerging markets. As a result, Mexico now attracts about 15 percent of total capital flows to emerging market economies.
E. Empirical Analysis
107. Empirical studies on dollarization and currency substitution have generally been based on simple models of relative currency demand. These models incorporate a mix of inflation, exchange rate expectations, and interest rate differentials as the main explanatory variables.11 We follow in this tradition, specifying a portfolio balance model in which assets are assumed to be imperfect substitutes and domestic residents can hold domestic currency, foreign currency denominated accounts in the domestic banking system, and domestic and foreign currency denominated bonds. The latter takes into account the increasing openness of the Mexican economy after NAFTA, and the increasing integration of its financial markets with those of the rest of North America.
108. We consider the following stylized money demand equations:12
where md and mf denote the real demand for deposits in Mexico at time t in local and foreign currency respectively; y is real income at time t; ρ is an unobserved time-varying “risk premium”on domestic assets, id and if represent nominal rates of return on domestic bonds and foreign bonds respectively. In this formulation, the relevant yield on domestic assets is defined as being net of risk, either in the form of default or exchange rate risk. Subtracting equation (2) from equation (1), we obtain the following equation for the log dollarization ratio (dt):
If we further assume that uncovered interest parity holds when allowance is made for the risk premium noted above such that
109. Following Ortiz (1983), we define the expected depreciation of the domestic currency as the inflation differential (πd-πf) between Mexico and the U.S. We also include in this stylized framework a variable vt, which we interpret to be a buffer for external shocks, defined as the ratio of foreign reserves to external short-term debt. The variable εt is a random disturbance term. The final specification for estimation is shown in equation (5):
110. As previously mentioned, the analysis uses a broader definition of dollarization than in previous studies. We measure dollarization as the ratio of foreign currency deposits of residents with domestic banks plus the deposits of Mexicans in U.S. banks, divided by the sum of M4 and the deposits of Mexicans in the U.S. banking system expressed in pesos. This helps to capture both foreign currency deposits in the domestic banking system and foreign currency held abroad.13
111. A brief description of the variables and the expected signs of their coefficients in the estimated model is provided below:
Real income. The effect of domestic income on the relative demand for dollar balances is ambiguous ex ante. Cuddington and Alami have argued that it is possible for the income variable to have a negative effect, depending on whether domestic portfolio considerations dominate transactions demand or not. Portfolio considerations dominate if the coefficient on income is negative.
Expected depreciation. According to the conventional portfolio model specification, an increase in expected rate of depreciation would lead to an increase in the demand for dollars as opposed to pesos, so that the coefficient would be positive. If there are factors that prevent agents from behaving as expected, for example, measures that restrict free movement between currencies or assets, the estimated coefficient could well be insignificant.14 The introduction of new measures restricting the holding of foreign currency deposits, as in 1996, could even result in a negative sign on the expected depreciation variable. Specifically, a forced shift into domestic currency assets would be expected to lower their equilibrium rate of return. In the context of equation (4), this would be reflected in an increase in the expected rate of depreciation. In other words, domestic currency holdings would be observed to rise at the same time as expected depreciation increased.
Return on dollar deposits. The coefficient captures portfolio shifts. An increase in the rate of return on foreign currency deposits is expected to increase the demand for foreign currency assets relative to domestic assets.
Vulnerability. We use the reserve coverage of short-term external debt to proxy the extent to which the economy may be vulnerable to external shocks.15 Assuming that a high level of reserves is taken as a signal that the central bank can respond effectively to shocks, thus imparting confidence in the domestic currency, we would expect the coefficient on this variable to be negative.
112. Our data cover the period 1990M1–2003M6. All data are in logarithms and were taken from the Bank of Mexico’s Indicadores Economicos (http://www.banxico.org.mx). It is well known that if time series are not stationary, the distributions of the conventional test statistics will be biased. The levels of all the series are nonstationary; differencing the data and performing unit root tests established that the variables could be characterized as I(1) processes.16
Cointegration and error correction
113. Since the variables in the model are I(1) and endogenous, we expect that the dollarization variable will be cointegrated with the variables on the right-hand side of equation (5). The long-term relationship corresponds to the cointegrating relationship(s), while the short-term dynamics—i.e. the error correction model—return the variables to equilibrium after a shock.17 The maximum likelihood technique of Johansen and Juselius (1990) is used to determine the rank (r) and identify a long-run dollarization relationship among the cointegrating vectors. The number of lags used in the VAR is based on the evidence provided by both likelihood ratio tests. When serial correlation of the residuals was present, a sufficient number of lags were introduced to eliminate it.
114. The null hypothesis of no cointegration was rejected using both the Λ-max (maximum Eigenvalue statistics) and trace tests, in favor of one cointegrating relationship. Both tests indicate one cointegrating vector at the 1 and 5 percent level. We then consider a dynamic error-correction model that takes the form of an autoregressive distributed lag to capture the short-run dynamics of the specification in equation (5). The regression results are presented in Table 1. The signs of the coefficients generally correspond to those implied by the portfolio balance, with the exception of that on exchange rate expectations which is negative suggesting the presence of an “identification problem” arising from the role of prudential regulations, as discussed above.18
Mexico: Estimation Results Using ARDL Procedure 1990–2003
Significant at the 5 percent level.
Mexico: Estimation Results Using ARDL Procedure 1990–2003
| Variable | Coefficient | T-Statistics | P-value | |
|---|---|---|---|---|
| Constant | 0.082 | 2.00 | * | 0.048 |
| Short-run dynamics | ||||
| d(Expected depreciation) | -0.024 | -0.86 | 0.389 | |
| d(Return foreign assets) | 0.019 | 0.22 | 0.812 | |
| d(Vulnerability) | -0.006 | -6.35 | * | 0.000 |
| d(Real income) | -0.017 | -0.64 | 0.527 | |
| Long-run dynamics | ||||
| Expected depreciation (-1) | -0.023 | -0.96 | 0.338 | |
| Return foreign on asset (-1) | 0.025 | 1.78 | 0.077 | |
| Vulnerability (-1) | -0.002 | -3.48 | * | 0.000 |
| Real income (-1) | 0.001 | 0.48 | 0.629 | |
| Dollarization ratio (-1) | -0.088 | -2.26 | * | 0.030 |
| Adjusted R-squared | 0.277 | … | … | |
| Durbin-Watson Statistics | 1.882 |
Significant at the 5 percent level.
Mexico: Estimation Results Using ARDL Procedure 1990–2003
| Variable | Coefficient | T-Statistics | P-value | |
|---|---|---|---|---|
| Constant | 0.082 | 2.00 | * | 0.048 |
| Short-run dynamics | ||||
| d(Expected depreciation) | -0.024 | -0.86 | 0.389 | |
| d(Return foreign assets) | 0.019 | 0.22 | 0.812 | |
| d(Vulnerability) | -0.006 | -6.35 | * | 0.000 |
| d(Real income) | -0.017 | -0.64 | 0.527 | |
| Long-run dynamics | ||||
| Expected depreciation (-1) | -0.023 | -0.96 | 0.338 | |
| Return foreign on asset (-1) | 0.025 | 1.78 | 0.077 | |
| Vulnerability (-1) | -0.002 | -3.48 | * | 0.000 |
| Real income (-1) | 0.001 | 0.48 | 0.629 | |
| Dollarization ratio (-1) | -0.088 | -2.26 | * | 0.030 |
| Adjusted R-squared | 0.277 | … | … | |
| Durbin-Watson Statistics | 1.882 |
Significant at the 5 percent level.
115. All contemporaneous short-run coefficients are statistically insignificant, with the exception of the vulnerability variable, which is significant and has the expected sign. The lagged values of return on foreign assets, the vulnerability variable, and the ECM coefficients are all significant in the long run. The explanatory power of the model is relatively low, however, suggesting that factors outside the model may have affected the dollarization ratio.
F. Conclusions
116. Previous attempts to estimate the magnitude and determinants of currency substitution in Latin America have presented evidence of hysteresis in the dollarization process. In the case of Mexico, however, dollarization has been trending down and displays little evidence of hysteresis. Our analysis suggests that the expected depreciation of the peso is significant in explaining dollarization in Mexico, but the sign on the coefficent is negative. We attribute this counter-intuitive outcome to the effectiveness of prudential regulations on foreign currency deposits since 1996. These regulations may have discouraged banks from operating such accounts, since the net return is marginal. On the other hand, a reduction in vulnerability to external shocks enhances the credibility of macroeconomic policies and induces a reduction in the level of dollarization in Mexico.
117. The results also suggest that prudential regulations could be effective in discouraging FCD holdings, without the distortionary effects of stringent direct measures to regulate FCDs. Mexico had to abandon the forced conversion of dollar accounts to domestic currency in 1987 as it substantially increased convertibility risk and the credibility of policy. This in turn led residents to demand sharply higher real interest rates to hold domestic currency accounts.19 Since then, Mexico has implemented policies that limit the incentives for ownership of FCD accounts while not directly limiting individuals’ capacity to hold such accounts in foreign countries. While the latter option, in effect, allows all Mexicans to open foreign currency accounts abroad through local banks in Mexico, it makes it unattractive for banks to want to do so since the net income from doing so is marginal. Complemented by the strict prudential regulations on banks, this has effectively allowed Mexico to de-dollarize its economy. Finally, our findings suggest that increased resilience of Mexican economy to external economic shocks played important roles in reducing dollarization.
References
Alami, T.H., 2001, “Variance Decomposition Analysis of the Demand for Foreign Money in Egypt,” Journal of Economic Studies, Vol. 28, pp. 122–135.
Berg, A., and Borensztein, E. R., 2000, “The Pros and Cons of Full Dollarization,” IMF Working Paper 00/50.
Calvo, G.A., and C.A. Vegh, 1992, “Currency Substitution in Developing Countries: An Introduction,” IMF Working Paper 92/40 (Washington: International Monetary Fund).
Clements, B., and G. Schwartz, 1992, “Currency Substitution: The Recent Experience of Bolivia,” IMF Working Paper 92/65 (Washington: International Monetary Fund).
Cuddington, J., 1983, “Currency Substitution, Capital Mobility, and Money Demand”, Journal of International Money and Finance, Vol. 2, pp. 111–133.
Dornbusch, R. and A. Reynoso, 1989, “Financial Factors in Economic Development,” NBER Working Paper No. 2889.
Guidotti, P. and C.A. Rodriguez, 1992, “Dollarization in Latin America: Gresham’s Law in Reverse,” Staff Papers, International Monetary Fund, Vol. 39, No. 3, pp. 518–544.
Gruben, W.C. and J.H. Welch, 1996, “Default Risk and Dollarization in Mexico,” Journal of Money, Credit and Banking, Vol. 28, No. 3.
Ortiz, G., 1983, “Currency Substitution in Mexico: The Dollarization Problem,” Journal of Money, Credit and Banking, Vol. 15, pp. 174–85.
Phillips, P.C.B., and P. Perron, 1988, “Testing for a Unit Root in Time Series Regression,” Biometrika 75, pp. 335–46.
Ramirez-Rojas, C.L., 1985, “Currency Substitution in Argentina, Mexico and Uruguay”, IMF Staff Papers Vol. 32, pp. 627–667.
Rojas-Suarez, L., 1992. “Currency Substitution and Inflation in Peru,” Revista de Análisis Económico, Vol. 7, No. 1, pp. 153–176.
Rogers, J.H., 1992, “The Currency Substitution Hypothesis and Relative Demand in Mexico and Canada,” Journal of Money, Credit and Banking, Vol. 24, No. 3.
Sahay, R. and C. A., Vegh, 1996. “Dollarization in Transition Economies: Evidence and Policy Implications”, in: Mizen, P. and Pentacost, E. J. (Eds.), Macroeconomics of International Currencies, Theory, Policy and Evidence, Edward Elgar, Brookfield.
Savastano, M.A., 1990, “Speculative Attacks and Currency Substitution under Managed Exchange Rate Regimes in Developing Countries,” mimeo University of California, Los Angeles.
Savastano, M.A., 1996, “Dollarization in Latin America: Recent Evidence and Some Policy Issues,” IMF Working Paper 96/04 (Washington: International Monetary Fund).
Sturzenegger, F., 1997, “Understanding the Welfare Implications of Currency Substitution,” Journal of Economic Dynamics and Control, Vol. 21, pp. 391–416.
Uribe, M., 1997, “Hysteresis in a Simple Model of Currency Substitution,” Journal of Monetary Economics, 40, pp. 185–202.
Yotopoulos, P.A. 1997, “Financial Crises and the Benefit of Mildly Repressed Exchange Rates,” Stanford University, Working Paper Series in Economics and Finance No. 202.
Prepared by E. Faal and N. Thacker.
Some transition economies, such as Estonia and Poland, have also experienced a decline in dollarization as their economies have stabilized.
Information on bank notes in circulation is not available, but is likely to be negligible as most transactions are in pesos.
In a rejoinder, Rogers tried to duplicate the Gruben and Welch analysis using their data set but was unable to reach similar conclusions.
M4 includes M2 plus domestic financial assets held by nonresidents plus deposits in branches and agencies of domestic banks abroad.
The current account deficit increased from 3.1 percent of GDP in 1978 to 5.8 percent of GDP in 1981 despite a sevenfold increase in oil receipts.
The previous change in regulations had been introduced in March 1977, when restrictions on holding FCDs were eased.
The country also experienced political turmoil and violence during this period.
Total resources include resources available in the domestic financial system (M4 plus bank’s external debt).
Estimates show that the net (including expected recovery of assets) fiscal costs of these various support programmes was about 20 per cent of GDP for 1995–97, half of which derived from the takeover of banks.
Rogers (1983), Cuddington (1983), and Alami (2001).
All variables are in logs, except interest rates.
This definition does not significantly alter the results compared with that using M4 alone.
As mentioned earlier, Rogers obtained a negative sign for this coefficient and interpreted this to imply the presence of convertibility risk.
This is consistent with the indicators used by the Fund to assess vulnerability of countries to economic crisis.
A useful discussion of unit root tests can be found in Perron (1989) and Phillips and Perron (1988).
One problem with the Johansen and Juselius procedure is that it is not able to exactly identify the parameters in the a and b matrices. Only if just one cointegrating vector is found can we make concrete conclusions about a unique long-run relationship between the variables.
We did not include a dummy variable for the 1996 changes in regulations in our analysis since we wanted to test for the reversal of the sharp dollarization episode of 1994–95. Including a dummy results in multiple cointegrating vectors.
