Dollarization persistence and individual heterogeneity

https://doi.org/10.1016/j.jimonfin.2010.05.001Get rights and content

Abstract

A salient feature of financial dollarization, arguably the one that causes most concern to policymakers, is its persistence: even after successful macroeconomic stabilizations, dollarization ratios often remain high. In this paper we argue that this persistence is connected to the fact that the participants in the dollar deposit market are fairly heterogenous, and so is the way they form their optimal currency portfolio. We develop a simple model when agents differ in their ability to process information, which turns out to be enough to generate persistence upon aggregation. We provide empirical evidence that is consistent with this claim.

Introduction

Even though dollarization is a relatively new research area, the experiences of many Latin American and transition economies during the 1990s has inspired a growing and rich body of literature (see, inter alia, De Nicoló et al., 2005, Levy-Yeyati, 2006, for recent reviews). Dollarization is normally associated with the partial substitution of the domestic currency by a foreign currency (the US dollar) as a store of value, as opposed to currency substitution which refers to the use of the foreign currency as a medium of exchange. In this paper, by dollarization we mean deposit dollarization, which eventually leads to credit dollarization and to the vulnerability of the financial system of highly dollarized countries. As stressed by Cook (2004), the efficacy of monetary policy in small open economies with flexible exchange rates is compromised by the negative balance sheet effects generated by dollarization. In this case, sudden real depreciations can have detrimental consequences on the economic activity by reducing the net worth of firms and generating adverse effects on investment. This situation gives a rationale for a “fear of floating” behavior of central banks (cf. Morón and Winkelried, 2005).

One of the most salient features of dollarization, and probably the one that causes most concern to policymakers, is its persistence. It is well documented that dollarization increases sharply during episodes of unduly macroeconomic instability and that it remains stubbornly high even long after successful stabilizations (cf. Guidotti and Rodríguez, 1992, Kamin and Ericsson, 2003). A top-of-mind explanation of this hysteresis is lack of confidence in domestic currency assets as the result of the traumas brought by past inflation, devaluations, banking crises, and so on. Thus, even if strong macroeconomic fundamentals can be observed in several highly dollarized countries (e.g., Peru in the early 2000s), persistence of dollarization may be explained by fears that the sound macroeconomic policies will someday be abandoned.

An alternative avenue to address this puzzle is to adapt the existing literature on currency substitution, which is based on adjustment costs or network externalities. Guidotti and Rodríguez, 1992, Sturzenegger, 1997 and Uribe (1997) develop models where the cost of using the dollar for transactions depends negatively on the aggregate currency substitution ratio, so once transactions get dollarized, there is no benefit to switch back to using domestic currency if others continue using dollars. However, a limitation of this approach for explaining financial dollarization persistence is that relies only on transaction costs, and does not take into account financial variables, fundamental determinants of dollarization. Moreover, these models rely heavily on a knowledge stock that drives the persistence (a “ratchet variable”). Thus, although they can explain upward trends in the depth of dollarization, they may not be satisfactory in tracking episodes of dedollariziation as this would require an implausible reduction in the knowledge stock.

On the other hand, Ize and Levy Yeyati (2003) derive a minimum variance portfolio (MVP), usually interpreted as the underlying equilibrium level of dollarization that depends on the relative volatility of inflation and real depreciation rates. Dollarization would persist even when inflation is low and stable insofar as the real depreciation volatility is smaller than that of inflation. The MVP approach has proven successful in explaining cross-sectional variation of dollarization level (cf. De Nicoló et al., 2005), but seems to be less satisfactory when it comes to explaining the evolution of dollarization ratios (see Morón, 2006). For instance, hysteresis is observed in several countries with high real exchange rate volatility (relative to inflation), such as Peru, Russia or Turkey. One explanation for this apparent contradiction with the portfolio approach is the difficulty in obtaining sound estimates of the variances and covariances that compose the MVP. On one hand, unobservable factors such as credibility effects or fears of a collapsing monetary regime can generate differences between observed and expected volatilities, the latter being the theoretical measurement of uncertainty in the MVP. On the other hand, in its simplest form, the MVP’s framework is static and hence depends on unconditional moments. Since persistence is inherently a dynamic phenomenon, our analytical framework allows the MVP to depend on conditional, time-varying moments.

Curiously, a fact that researchers have apparently overlooked is the very nature of the participants of the dollar deposit market in dollarized economies: depositors are extremely heterogenous, ranging from large entrepreneurs to small firms to non-profit organizations and to individuals, rich and not-so-wealthy.1 Participation costs in the dollar market are virtually nil due to liberalization, deregulation and, importantly, due to the emergence of informal currency traders – known as cambistas in Latin American countries and as obmenschiki in Russia – which benefit from buying and selling dollars with tighter markups than those in the banking sector.2 A typical cambista or obmenschik would hold a limited amount of money for business (say, between US$2000 and US$5000) as he seeks to meet the demand for dollars of individuals or small firms, normally unwilling to pay the higher bank premium to get their savings dollarized.3 As a result, participation becomes independent of the scale of the transaction and hence widespread.

The aim of this paper is to draw the attention to the fact that heterogeneity of depositors can help explaining the persistence of financial dollarization. As first pointed out by Granger (1980), differences in individual dynamics could lead to aggregate persistence. Thus, as it is reasonable to expect that the dynamics of the optimal currency portfolio of a financial expert differ from that of a blacksmith, a persistent aggregate dollarization ratio arises naturally. There are of course various differences between a financial expert and a blacksmith, but provided that both access the dollar deposit market almost freely, the relevant difference in our analysis is their ability to process information and, therefore, to take informed portfolio decisions.4

The rest of the paper is organized as follows. Section 2 motivates the analysis, more precisely the interplay between individual heterogeneity and aggregate persistence, with a brief account of the Peruvian and Polish experiences. It also gives an idea of how the dollar deposit market in representative countries is shared among various types of deposit holders. Section 3 develops a stylized model where agents face noisy information and differ in their ability to forecast when taking portfolio decisions. An important result is that it is optimal for agents to be cautious when modifying the currency composition of their deposits as there is uncertainty on the quality of the data agents receive. The caution is reflected in portfolios that may adjust in a relatively slow fashion. We then show that upon aggregation of the individual dollarization decisions it is possible to generate a persistent economy-wide dollarization ratio.5 Section 4 explores whether the results of the theoretical model are consistent with the dynamics of aggregate data of Mexico, Peru, Poland, Turkey and Uruguay. The empirical results suggest that the underlying distributions of “forecasting abilities” behind the aggregate figures would be very spread and often skewed. We regard this result as consistent with the idea of financial experts sharing the dollar market with blacksmiths that save in dollars. In Section 5 we discuss some caveats to the analysis, and Section 6 concludes, gives policy recommendations and suggests avenues for future research. Derivations and complementary results are shown in three appendices.

Section snippets

Two illustrative cases

As stressed in Savastano (1996), dollarization emerges progressively in response to macroeconomic instability, particularly high levels of inflation, showing a well-defined pattern: first, agents replace domestic currency as reserve of value, holding usually dollars outside the financial system (“under the mattress”). Then, the dollar is used in some transactions, typically involving real estates and durable goods, and eventually some prices are set in dollars. Most governments later on allow

A simple model

The main insight of our analysis is that to the extent that the information agents need to determine their portfolios changes over time, and that they may receive these news imperfectly (i.e., with noise), the different capabilities in processing information play a key role on the aggregate dynamics of dollarization ratios. Next, we use a simple framework to show how the combination of imperfect, noisy information on real returns of foreign assets, and the heterogeneity among market

Empirical exploration

This section explores whether the dynamics of the dollarization ratio can be regarded as coming from the aggregation of heterogeneous depositors. For this undertaking we use information from a sample of selected countries that differ in the degree of financial development, size and economic structure and also differ in their dollarization experiences. In this way we intend to provide a robustness check of our model’s main predictions to country-specific features. It is imperative to highlight

On the role of learning

An alternative way to rationalize the fact that individuals are heterogeneous in their forecast of Rt is to assume that they cannot perfectly observe the true process that governs the evolution of Rt. For instance, because they do not know the exact value of α in Eq. (3). In this case, individuals should form priors on the value of this parameter in order to forecast Rt and to make their portfolio choices. Agents may have different priors on α, but they can update them as new information on Rt

Concluding remarks

In countries with some degree of dollarization, participation in the dollar deposit market has become massive. Financial deregulation, liberalization, innovation and informal currency markets have allowed a very heterogenous group of agents – from a large firm that uses state-of-art portfolio management techniques to uninformed individuals who base their portfolio decisions simply on their own experience and limited information – to participate in the same market. This paper shows that such a

References (30)

  • P. Zaffaroni

    Contemporaneous aggregation of linear dynamic models in large economies

    Journal of Econometrics

    (May 2004)
  • P.R. Agénor et al.

    Macroeconomic management with informal financial markets

    International Journal of Finance and Economics

    (April 1996)
  • J. Arifovic

    The behavior of the exchange rate in the genetic algorithm and experimental economics

    Journal of Political Economy

    (April 1996)
  • C. Arteta

    Exchange rate regimes and financial dollarization: Does flexibility reduce currency mismatches in bank intermediation?

    Topics in Macroeconomics

    (2005)
  • T. Beck et al.

    A new database on the structure and development of the financial sector

    World Bank Economic Review

    (2000)
  • Cited by (9)

    • Policies for transactional de-dollarization: A laboratory study

      2022, Journal of Economic Behavior and Organization
      Citation Excerpt :

      Similarly, Vieira et al. (2012) characterize dollarization empirically (using a cross-country estimation) and document it as an insurance response in an environment of high domestic debt, inflation, and default risk. Winkelried and Castillo (2010) model an emerging country’s economy and argue that dollarization arises as a result of heterogeneous information processing among agents in that economy. Finally, Imam and Corrales (2019) estimated a panel data model to explain the financial dollarization of households and firms across countries.

    • Taming dollarization hysteresis: Evidence from post-socialist countries

      2023, Handbook on Critical Political Economy and Public Policy
    • Asymmetries in Volatility: An Empirical Study for the Peruvian Stock and Forex Markets

      2019, Review of Pacific Basin Financial Markets and Policies
    View all citing articles on Scopus

    We would like to thank Kosuke Aoki for his valuable advice. We are also grateful to Bernardo Guimaraes, Carlos Arteta, an anonymous referee and seminar participants at the 2004 LACEA meeting, the Central Bank of Peru and the University of Cambridge for helpful comments on earlier versions of this paper. The views expressed in this article are those of the authors and do not necessarily represent those of the Central Bank of Peru. Diego Winkelried gratefully acknowledges financial assistance from the ORS award and the Gates Cambridge Trust.

    View full text