آسیب پذیری چین در بحران ارز: رویکرد سیگنال های AKLR

In this paper we use the Kaminsky–Lizondo–Reinhart (KLR) [Kaminsky, G., Lizondo, S., Reinhart, C., 1998. Leading Indicators of Currency Crises. International Monetary Fund Staff Papers 45, 1–48.] approach to conduct an ex-post study of the probabilities of China suffering a currency crisis during the period of January 1991 to December 2004. Two high-probability periods are identified: July 1992–July 1993 and August 1998–May 1999. The first period correctly predicts China's 1994 devaluation. The second period predicts currency devaluation in the aftermath of the Asian crisis, which did not occur. The results of the model indicate that the fundamentals were weak enough for China to experience contagion of the Asian crisis, and raise the question of the possible role of China's institutional arrangements in preventing the crisis. The paper further analyzes the economic fundamentals of China that drive the high probability of crises, and provides some suggestions for further reform.

China's economic transformation has been one of the major development successes of the end of the twentieth century. Since the start of the economic reforms in 1978, China has outperformed any other country in the world in terms of economic growth, taking 400 billion people out of poverty in the process. China has become one of the top six largest economies in terms of output produced (when output is measured at market exchange rates), the fourth world trader, and the largest recipient of foreign direct investment (Prasad & Rumbaugh, 2004). Despite all this success, the Chinese economy still has some structural weaknesses that may undermine its potential for future growth. In particular, China's financial system is weak: financial intermediation is mainly dominated by state banks, which have high proportions of non-performing loans in their asset portfolios. Furthermore, the currency is non-convertible, and capital controls substantially restrict the ability of Chinese citizens to invest abroad. This institutional structure is in the process of being reformed by the Chinese government, which has expressed its willingness to liberalize its capital account. China is also under pressure from the United States and other economies to revaluate its currency and potentially liberalize its currency system. Some economists see this move as problematic and fear a reproduction in China of the circumstances that led to the Asian financial crisis of 1997 (Goldstein & Lardy, 2006). Given the strong link between China and other Asian economies, destabilization in China could be readily spread over the region. The East Asian crisis of 1997 was prompted by Thailand's devaluation of its currency in mid-1997 and it rapidly spread to other East Asian economies, which either had to devaluate their currencies or lost important amounts of foreign reserves defending them. Even though the East Asian countries affected by the crisis differed in terms of their economic policies and economic fundamentals, they had some common features that made them vulnerable to a currency crisis. In particular, most economists agree by now that the crisis was the result of a combination of weak financial systems with high exposure to short-term foreign debt (Goldstein, 1998, Lee, 2003 and Radelett and Sachs, 1998, among others). China was one of the few East Asian emerging economies that were spared from contagion of the currency crisis. There has been a lot of speculation on how fragile the Chinese economy was during the Asian Crisis and the possible role that its exchange rate regime may have played at insulating the country from the crisis. Some authors attribute China's ability to resist contagion to its better fundamentals (Lan, 2002); whereas other authors claim that the non-convertibility of the RMB and capital controls had a crucial role in the insulation (see, for instance, Lardy, 2003 and Lee, 2003). In order to analyze the level of vulnerability of a country to currency crisis at any given period, a method is needed that is able to identify a set of leading indicators that present abnormal behavior prior to a currency crisis and that gives a measure of the likelihood of a crisis occurring, given the behavior of those indicators. This is precisely the objective of “early warning system” approaches. Even though these methods are designed as forecasting devices, they are also useful, when applied to a single country, in determining the ex-post likelihood of that country suffering a currency crisis at any given period (Edison, 2003). In this paper we use an early warning system developed by Kaminsky, Lizondo, and Reinhart (1998) (the KLR signals approach) to determine, ex-post, the probabilities of currency crises for China for the period of January 1991 to December 2004. Two periods of high probability of crisis are identified: July 1992–July 1993 and August 1998–May 1999.1 The first period correctly predicts the exchange rate realignment of 1994, when China unified its official and swap-market exchange rates, resulting in a 50% devaluation of the renminbi (RMB). The second period coincides with the aftermath of the East Asian crisis. Even though no devaluation occurred in China during this period, the high probability of crisis indicates that, according to the KLR method, China's fundamentals during the Asian crisis were weak enough to make the country a candidate for contagion. The paper further analyzes the economic fundamentals underlying the “crisis signaling” periods. We find that the fundamentals signaling the crisis in both periods are radically different, except for an increase in the M2 multiplier, which appears on both periods. For the 1992–1993 periods the signaling variables are a rise in M2, a decrease in reserves, and an appreciation of the real exchange rate. For the 1998–1999 periods the indicators signaling a crisis are the real interest rate differential, domestic credit relative to GDP, exports, and terms of trade. An analysis of the Chinese economy at that period reveals that the decrease in exports is a direct effect of the recession that the Asian Crisis brought to the region (an important part of Chinese trade at the time was with other East Asian countries). The increase in domestic credit was driven by the government massive investment in infrastructure in order to keep the Chinese growth rate at its target level of 8%. This suggests that the government used expansionary fiscal policy in order to face the challenges posed by the Asian crisis, instead of devaluating the currency. The objective of this paper is twofold. First, by analyzing the Chinese economy under the KLR method we develop a better understanding of the vulnerabilities of China during the Asian crisis and the reasons why China was able to avoid contagion. Second, we shed some light on the use of the KLR method for countries with non-convertible currencies and capital controls. To the extent that currency inconvertibility and capital controls reduce the possibility of sudden and massive capital flights, weak fundamentals may not necessarily lead to currency devaluation. We argue that for such countries, the KLR method is a useful tool to detect structural weaknesses in the economy that would lead to exchange rate pressure in a liberalized economy. Detecting these weaknesses is especially important for economies that are in the process of reform and liberalization of their financial and exchange rate systems, since they may increase their vulnerability to currency crises once the controls are lifted. There is a growing literature based on the KLR signals approach. First developed by Kaminsky et al. (1998) the method determines a set of relevant indicators which present abnormal behavior prior to a crisis. The indicators are determined by observing the experiences of a set of countries during periods of currency crises (sample-based method). For each of the indicators, the method determines a threshold value. A variable that surpasses its threshold value is assumed to issue a signal. The probability of a crisis occurring in the next 24-month period is determined by the nature and amount of indicators issuing signals at a given point in time. Berg and Pattillo (1999) evaluate the predictive power of the KLR approach. Edison (2003) augments the KLR with a few additional explanatory variables and expands the sample of countries used in determining the indicators and thresholds by including more developing countries. Edison (2003) constructs a new set of threshold levels and probabilities of crises based on his expanded sample. Edison also evaluates how this approach can be applied to an individual country, and finds that the model provides some useful information about a country's vulnerability to a crisis. Kaminsky (1998) introduces a composite crisis indicator that she uses to determine the probability of crisis in each period of time. She then applies the model to some out of sample countries to determine, ex-post, whether the method is able to correctly predict their currency devaluations during the Asian crisis. She finds that the method correctly predicts the currency crisis in Thailand, Malaysia and the Philippines, but it fails to predict the crisis in Indonesia. Alvarez-Plata and Schrooten (2004) apply the method to the analysis of the 2002 currency crisis in Argentina. A few remarks should be made about the application of the KLR method to the case of China. First, it is well known that the quality of the Chinese data is questionable (see, for example, Young, 2003). Regarding output data, local authorities may over-report industrial output in order to keep with government targets. Second, there is the question of the information revealed by financial data in a country with vast government controls on the financial and exchange rate system. Lending rates are controlled by the government and may not reflect market conditions. Capital controls reduce people's ability to diversify portfolios and the fact that banks are government owned implies an implicit guarantee on deposits. Therefore, deposit rates may be less responsive to financial fragility than in free market economies (Fernald and Babson, 1999 and Prasad and Rumbaugh, 2004). The limitations on the information contained in the financial data discussed above reduce the ability of the KLR model to correctly predict crises in economies like China. The indicators are either government determined (interest rates) or are less responsive to changes in the economy (deposit rates). Furthermore, massive capital flights are not likely to happen, which may increase the set of policy tools that the government can use to face the deterioration of fundamentals in the economy. In summary, for an economy like China we should expect less “crisis signaling” periods when applying the KLR method. Even when the KLR method delivers a high probability of crisis occurring in a given period, we should expect lower probabilities of observing an actual crisis occurring. Therefore, when applied to economies like China, we should view the KLR model more like a detector of weak fundamentals and less like a prediction device. Detecting weak fundamentals is especially valuable for countries that are undergoing financial liberalization, since they can be destabilizing once the reforms are implemented. The index of exchange market pressure also has a different meaning in a country with capital controls and fixed exchange rates. Given that massive capital flights are unlikely to occur, strong variations in reserves are very unlikely to occur in short periods of time. Therefore, the index is less of a measure of the reaction of investors to fundamentals, and more a measure of fundamentals per se. Therefore, the type of “crises” captured by the index will be either situations where the government chooses devaluation as a policy tool, or situations where foreign reserves are depleted for causes other than investors' behavior. Finally, the usual caveats regarding the weaknesses and limitations of using the KLR model to identify crises episodes apply here. First, the KLR method is sample dependent. The set of relevant indicators, as well as the threshold levels that determine whether the indicator is issuing a signal depend on the base data sample. Expansions of the base sample to include new countries or more time periods may result in the disappearance (or appearance) of some crisis episodes (Edison, 2003 reports that these changes in dating schemes may occur after a big crisis is included into the base sample). Second, the method often predicts crises that do not occur. These false positives are in part due to the fact that crisis episodes do not occur that often in the data and, therefore, most observations in the base sample are tranquil periods. Third, the KLR method does not account for possible endogeneities between the method's predictions and the behavior of the government and other economic agents. That is, the method assumes that some specific indicators systematically present abnormal behavior prior to a crisis, but it does not incorporate the reaction of the government and other economic agents to this type of knowledge. In particular, after observing the abnormal behavior of those indicators, the government may choose to implement policies that reduce the risk of a crisis occurring, or investors may precipitate a crisis if they know that the method predicts that one could occur in the near future. Fourth, the KLR method abstracts from country-specific institutions, which may be important in determining the economy's response to the abnormal behavior of other indicators. Despite its limitations, the model is still a useful diagnostic tool that helps to identify periods of weak fundamentals where a country was vulnerable to currency crisis. It is this feature of the KLR method that we exploit in this paper. The rest of the paper is organized as follows: Section 2 briefly describes the Chinese economy in the period of study. Section 3 provides a description of the methodology. Section 4 presents the results and Section 5 concludes and provides some suggestions for further reform in China.

In this paper we use the KLR signals approach to conduct an ex-post study of the probabilities of currency crises for China for the period of January 1991–December 2004. The method correctly predicts the exchange rate realignment of 1994, and signals the aftermath of the Asian Crisis, August 1998 to May 1999 as another period with high probability of crisis. The fact that China did not experience an actual currency crisis in that period suggests a potential role for China's capital controls and non-convertibility regime in insulating China from the Asian crisis. The results of this paper have some implications for future policy. The underlying economic weaknesses that China had during the Asian crisis, namely a fragile financial system, are still in place today in the Chinese economy. First, it is clear that a largely insolvent banking sector already exists in China, which leaves the economy vulnerable to future financial crisis. Hence, the creation of a modern banking system with adequate supervision and regulation is essential for China today. Second, the majority of the state-owned enterprises are insolvent. Moreover, their dependence on state-owned banks has left the banking sector with large non-performing loans. Further SOE reform is needed if China is to keep its current growth. Third, our analysis suggests that China's exchange rate regime had a potential role in sparing the country from the Asian crisis. If this is indeed the case, a liberalization of the capital market or a sudden realignment of the exchange rate peg without addressing the problems of the financial system may make China more vulnerable to future currency crises. A theoretical model is needed in order to fully understand the links between the exchange rate regime, the fragility of the financial system, and the vulnerability to currency crises. Constructing such a model is left for further research. This paper also derives some implications for the use of the KLR approach. First, by applying it to a country with capital controls and non-convertibility of the currency it sheds some light on the interpretation of the results of the KLR method. In particular, based on this application to the case of China, financial fragility may not lead to currency crises in economies where short term massive capital flights are unlikely. Drops in reserves, on the other hand, may have effects similar to the ones experienced by more liberalized economies. This observation suggests a possible role for institutional variables in early warning approaches.