Financial crises, liberalization, and government size.

• Author: Chu, Kam Hon

There is a broad consensus among economists that financial crises are costly, as evidenced by the Asian currency crisis in 1997 and other systemic crises during the 1990s (Hawkins 1999; Klingebiel and Laeven 2002). However, there is little agreement on the cause of financial fragility, not to mention the policy prescriptions for financial stability. A perennial heated controversy is the role of government versus the market in promoting and maintaining financial stability. Against the background of frequent outbreaks of financial crises following the global trend of financial liberalization over the past quarter of a century, many economists have pointed to financial liberalization as an important source of financial instability. For example, Jomo (1998) argues that the Malaysian crisis in 1997 was due to financial liberalization rather than excessive regulation. Empirically, such a view is to some extent supported by certain studies that show that financial liberalization has induced excessive risk taking by financial institutions and ultimately precipitated financial crises (Demirguc-Kunt and Detragiache 2001, 2005; Noy 2004).

Since the Asian crisis, many economists and policymakers have called for stronger regulation of financial markets. A popular view is that global financial markets are now beyond the control of governments and that financial crises are the consequence (Strange 1998, 2002). Adherents of that view call for tighter state control over financial markets.

The literature on financial instability has been expanding rapidly over the last decade, and the literature on the relationship between the state and the market has an even longer history. In this article, I employ categorical data analysis techniques to examine whether there are statistically significant associations among financial crises, liberalization, and government size. If there are associations, I proceed to ask: Does financial liberalization have a significant impact on financial stability? And, more important, does a large government avert financial crises? Does the impact of one factor, say financial liberalization on financial stability depend on the other factor (i.e., the size of government), and vice versa? The answer to that question can potentially shed light on the role of the state versus the market in promoting and maintaining financial stability. The main conclusion of this article is that the popular belief that financial crises are due to governments being outgrown by markets cannot be substantiated by the data.

Financial Deregulation, Government Size, and Financial Stability

To examine the relationship between financial deregulation, government size, and financial instability during the past three decades, I employ contingency table analysis and log-linear models. Financial instability is regarded as a response variable, whereas financial deregulation and government size are treated as factors or explanatory variables.

A country is defined as having experienced financial instability if it had at least one systemic banking crisis or borderline case during the period under study. Caprio and Klingebiel (1996, 1999, 2000) documented countries with systemic crises or borderline cases during the past three decades. For countries that experienced more than one financial crisis during the period under study, the following criteria apply: (1) when there is a systemic crisis and a borderline case for a country the systemic crisis will be selected; (2) when financial crises for a country are the same in terms of severity the choice will depend on the availability of data on financial deregulation and government size; and (3) when financial crises fall into the same category of severity and data availability is not a problem the country's latest financial crisis will be chosen.

After a country's status regarding financial instability has been determined, I consider the changes in that country's government size and in the extent of its financial regulation over a long period--say, a decade or more, depending on data availability--prior to the outbreak of the financial crisis. This approach assumes that financial stability responds to the state of financial deregulation and government size, and that policy or regime changes take time before they fully exert their impacts on the financial sector. Data on government size and financial regulation are from the dataset compiled by James Gwartney and Robert Lawson (2005a, 2005b). (1) Since these authors report their data on a five-year basis starting in 1970, the data on changes in government size and financial regulation used in this article are in most cases not the changes over the decade prior to the outbreak of a financial crisis. For instance, Hong Kong experienced a borderline case of financial crisis during 1982-86, and hence changes in government size and in financial regulation over the period 1970-80 are used in this study because data for 1971-81 are not reported in Gwartney and Lawson's dataset.

In a small number of cases, the time span is shortened to less than a decade because the relevant data over a longer period are unavailable. Such variations in the data are expected to have little, if any, distortions in reflecting the long-term trends in government size and financial regulation among the sample countries. The reason is that our empirical analysis is based on categorical data rather than on actual changes in the numerical values of those variables.

From Gwartney and Lawson's dataset, general government consumption expenditure as a percentage of total consumption is used as a proxy for government size. (2) One may argue that this proxy understates the actual size or scope of government because it omits transfer payments, which have been growing dramatically in many countries. But as long as government consumption and transfer payments follow the same long-run growth trend, this proxy should not result in any serious bias because our empirical analysis is based on categorical data. All other things equal, higher government consumption expenditure as a percentage of total consumption represents a larger government sector. It also reflects a more powerful or influential government because political choice is substituted for private choice. For those countries that have experienced financial crises, changes in the size of government before the outbreaks of financial crises are computed. Based on the calculated changes, countries are classified as having either smaller or larger government sectors.

Gwartney and Lawson's dataset 'also includes three indexes reflecting the extent of financial regulation or restrictions. (3) These include freedom to own foreign currency bank accounts domestically and abroad, international capital market controls, and credit market regulations. For each country I take the average of these indexes as a proxy for the overall level of financial regulation and then compute the corresponding changes in financial regulation over time. Similarly, countries are classified as having either deregulated or controlled their financial markets.

Finally, for countries that did not experience any financial crisis during the period under study, changes in government size and financial regulation over the decade 1980-90 are computed in a similar way for the purpose of statistical analysis and comparison.

The Empirical Results

Based on the available data, our sample includes 113 countries. Those countries' experiences regarding financial crises, government growth, and financial deregulation are summarized in Table 1. Of the 67 countries that experienced financial crises, 35 had deregulated their financial systems and had larger government sectors at the same time during the decade or so prior to the outbreaks of their financial crises. Meanwhile, 15 countries with the same characteristics did not experience financial crises. Other combinations can be read from Table 1 in a similar way.

To test if financial crises, the size of government, and deregulation are mutually independent, a [chi square] goodness-of-fit test is performed. The value of the [chi square] statistic is 8.6, which is not statistically significant at the conventional 5 percent level but is significant at the 10 percent level. (4) This finding...