1. Introduction介紹
 

如果資本熱潮會影響其突然停止的可能性，是有可能，突然停止可能影響資本熱潮發生類似的方式？

在我們的文章中，我們希望把重點放在資本的繁榮和突然停止，而不是一維的因果關系之間的相互作用。也就是說，我們將測試走向資本吊桿突然停止的預測功率;專注于行為金融因素，而不是在解釋新興市場的資本流動波動的基本面;考慮到非外商直接投資作為資本流動變化的因素;

我們將首先在一個數量的方式定義資本動臂和斗突然停止。然后，我們會嘗試找出根本的因素，如一束控制變量，如GDP增長率，經常賬戶赤字，外債，儲備，利率，匯率等，最后，我們將采取使用概率分析的測試互動資本動臂和斗突然停止，之間看到突然停止的發生是否會增加在未來時期資本繁榮的可能性
Our research is mainly based on De Bondt and Thaler(1985)’s theory, which studied the overreaction in stock market, and applies it to capital flow market in emerging countries. If capital boom can affect the possibility of having sudden stop, is it possible that sudden stop can influence the occurrence of capital boom the similar way?

In our paper, we would like to focus on the mutual interaction between capital booms and sudden stops instead of one-dimension cause-effect relationship. That is, we will test the predicting power of Sudden Stops towards Capital Booms; Focus on behavioral finance factors rather than fundamentals in explaining fluctuations of capital flows in emerging markets; Taking into account Non-FDI as a factor for change of capital flows;

We will firstly define Capital Booms and Sudden Stops in a quantities way. Then, we will try to figure out fundamental factors as bunch of control variables, such as GDP growth, current account deficit, external debt, reserves, interest rate, exchange rate, etc. Finally we will take use of Probit analysis to test the interaction between Capital Booms and Sudden Stops, and see whether the occurrence of sudden stops will increase the probability of capital booms in the next periods

2. Literature Review文獻綜述
 

This paper will learn from De Bondt and Thaler(1985)’s theory, which plans to study the overreaction in stock market, and applies it to capital flow market in emerging markets. De Bondt and Thaler (1985) brought on the Overreaction Hypothesis, in which they found an upward overreaction subsequently calls forth a dramatic downward adjustment in the stock market. Guidotti et al (2004) applied this theory to capital markets and defined Capital Booms and Sudden Stops, as those when capital flows fluctuate by more than a standard deviation of their average variation during the sample period. Manuel Agosin (2011) extended on Capital Booms Theory and found the best predictor of sudden stop is the preceding capital boom. The probability of a country undergoing sudden stop increases considerably with the length of boom.#p#分頁標題#e#

In our paper, we are going to conduct probit analysis based on over 100 emerging economies from 2005 to 2011, the result will show whether capital boom is an important predictor for sudden stop in the next period. At the same time, we will use a package of potential economic fundamentals at control variables to find out whether the probability of sudden stop is strongly predicted by previous capital boom. 

This paper contributes in three ways. Firstly, this paper emphasizes the power of behavior finance in predicting capital flow, contradicting the previous study which focused on economic fundamentals. In addition, the result reminds policy-makers the potential volatility risk when integrating into international financial market. Last but not least, we take into account NON-FDI as a potential factor for capital flow volatility, which has shorter horizon than FDI and is susceptible to reversion as investors acquire new information.

3. Hypothesis假設

Since we consider sudden stops as signals for underinvestment of a specific capital market, we expect a shot-term reversion, or to say, capital boom, will happen in the next period. That is, the occurrence of sudden stop will increase the probability of capital boom in the next periods. The correlation coefficient between capital boom and sudden stop should be positive.

If we get meaningful result, it will contribute in two ways. On one hand, it will tell us the importance of behavior finance in predicting capital flow, while most studies only focus on economic fundamentals. On the other hand, the result will remind policy-makers in emerging markets the potential volatility risk when integrating into international financial market.

4. Data and Methodology數據和方法

4.1Definition of Main variables

4.1.1 Capital Boom

If the financial account from the balance of payments is one standard deviation above its mean, we consider a capital boom exists. We define the capital boom indicator of a country i at year t 〖FF〗_it like below.

〖FF〗_it={█( 1        if F_it>(F_1 ) ?+σ_(F_i )@0                  otherwise)┤    (1)

F_it: The balance of the financial account of country i in year t (measured as current U.S.dollarsdeflated by the U.S.CPI).

(F_1 ) ?: The mean of F_it for the entire period.

σ_(F_i ): The standard deviation of F_it.

We prefer to use the level of capital flows instead of annual change of capital flows because we hope to test for the impact of large levels rather than large changes in capital flows.Besides, Calvoet al (2004) and Guidotti et al (2004) also require that the ratio of capital flow to GDP in the corresponding period is more than 5% when defining 〖FF〗_it to emphasize the effects of volatility on economy. We ignore this condition because we find that capital inflow in emerging markets is just in its early stage and as a result the ratio of capital flow to GDP cannot reach a relatively high level. #p#分頁標題#e#

4.1.2 Sudden Stop

Ifthe annual decline in the financial account of the balance of payments is one standard deviation below its mean, we consider a sudden stop exists. We define the capital boom indicator of a country i at year t 〖SS〗_it like below.

〖SS〗_it={█( 1        if F_it>(F_1 ) ?-σ_(F_i )@0                  otherwise)┤    (2)

The notations here are the same as in 4.1.1.

The phenomenon that a sudden stop in a particular year is followed by another sudden stop the following year may occur several times, we only identify the first contractionto be a sudden stop, because we are interested in identifying the start of a contradiction in capital inflow.

4.2Variable notations and calculations

The notations and calculations of variables used in our model are listed in the table below.

Variables Notations Calculations

Capital flow F Central bank financial account balance divided by U.S. CPI

Growth rate of GDP GDP Change in GDP divided by last-period GDP

The ratio of external debt to exports EDGDP The sum of public debt and private debt divided by GDP

The current account deficit as a share of GDP CAD Current account balance divided by GDP

Non-FDI flows in the financial account NONFDI Portfolio values divided by GDP

Change in the terms of trade TT The changing rate of the ratio of exports to imports

real growth rate of exchange rate RER Effective exchange growth rate(2005=100)

foreign real interest rate Rf U.S. 3-month interbank offer rate divided by CPI

Domestic real interest rate Rd Domestic deposits rate divided by CPI#p#分頁標題#e#

GDP growth rate of G7 countries G7gr The average of GDP growth rate of G7 countries

Degree of economic openness Openness The sum of exports and imports divided by GDP

The ratio of foreign reserves to GDP Reserves Foreign reserves divided by GDP

the ratio of M2 to GDP M2GDP M2 divided by GDP

4.3 Regression models

After defining the main variables, we will conduct probability analysis to test how capital boom affects the probability of sudden stop.

Wooldridge (2005) finds that average marginal effects of such discrete binary variables are not affected by form of models. So, for simplicity, we suppose the probability that capital boom will continue next year follows the standard normal distribution. We will adopt the discrete binary variable selecting model——Probit analysis——to analyze the sample data.

Let ∅ be the standard normal distributionN(0,1), we estimate a panel probit with heterogeneous unobserved effects. 

Pr(〖FF〗_it=1/〖FF〗_(it-1),〖NONFDI〗_it,X_(it-1),c_i)=∅(y_f 〖FF〗_(it-1)+y_nf 〖NONFDI〗_it+X_(it-1) β+c_i           (3)

Pr(〖SF〗_it=1/〖FF〗_(it-1),〖NONFDI〗_it,X_(it-1),c_i)=∅(y_f 〖FF〗_(it-1)+y_nf 〖NONFDI〗_it+X_(it-1) β+c_i           (4)

NONFDI: non-FDI flows in the financial account as a share of GDP.

X: a matrix of control variables.

c: error term.

Here lag-1 variable is chosen to reflect the lag effects existing in capital inflow. Short-term overreaction will result in overreaction of the market next period, and thus leading to a self-created continuous overreaction cycle. It is consistent with the widely adopted method “chartist strategy” by the global institutional investors.

Guidotti et al (2004) finds that the investment cycle of FDI is a long time and capital won’t flow in or flow out by a large amount in a short time. So, FDI are seldom chosen as the indicator of short-term “Herd Effect”, and we instead choose NONFDI as the explanatory variable.

According to Wooldridge (2002), we estimate (3) and (4) separately by the way of stochastic effect method. The average marginal effect can be calculated as below (For simplicity, the above equation omits all the time-series and cross-sectional subscripts.):

(∂E[P(y=1/X,c)/c])/(∂X_j )=(∂φ(Xβ/σ))/(∂X_j )=β_j/σ φ(Xβ/σ)        (5)#p#分頁標題#e#

y: endogenous variable FF or SS

X: matrix, representing all the other variables on the right hand side of equation (3) and (4).

c: follows conditional distributions c/x~N(0,τ^2), then 〖〖σ=(1+τ〗^2)〗^(1/2)

φ: standard normal distribution density function.

Since equation (3) is dynamic panel data, we give other assumptions. Referring to Wooldridge (2002 and 2005), we suppose:

h(c_i/y_i0,X_i,δ)=φ+δ_0 y_i0+(X_1 ) ?δ+a_i          (6)

a_i~N(0,〖σ_a〗^2).

i: cross-sectional units.

y_i0: The beginning value of the variable.

(X_1 ) ?: a matrix consisted of average of explanatory variables. 

After giving so many preparation equations, we derive our final regression equation:

〖y_it〗^*=y_f y_(it-1)+X_(it-1) β+φ+δ_0 y_i0+(X_1 ) ?δ+a_i+u_it

whereu_it~N(0,1)

4.4 Data statistics

4.4.1 Data collection

We are going to choose data from about 100 developing countries in the year 2005 to 2011, which is panel data. They are accessible through IMF and WDI. As for the regression, we are going to use panel probit regression by Stata. The number of all variables is going to be 14. And approximately, we will get 10000 items of data. The variables are as followed:

FF SS EDGDP NONFDI CAD Reserves M2GDP

GDP Rd TT Openness RER G7gr Rf

4.4.2 Countries with overreaction

The chart below states all effective FF and their corresponding countries. By previous definition of FF, we get 150 samples of capital over inflow. There are 29 in 2005, 34 in 2006, 12 in 2007, 9 in 2008 and 28 in 2009,19 in 2010 and 19 in 2011.As we can see, the frequency in higher in 2005, 2006 and 2009.If we consider about the international finance environment these years, this result is easy to understand.2005 and 2006 are good years for finance, and all developing countries were eager to import foreign capitals. And 2009 is the year after financial crisis, the capitals returned from the disaster. In the whole, the FF is reasonably distributed. #p#分頁標題#e#

4.4.3Reverse

The following chart gives all SS samples and their corresponding countries.By the definition above, we get 139 samples of capital over inflow. Among them, there are 6 in 2005, 3 in 2006, 19 in 2007, 42 in 2008, 14 in 2009, 21 in 2010 and 34 in 2011. The frequencies in 2008, 2010 and 2011 are higher. In 2008, due to the global recession in finance, all countries showed a downturn in capital flows, and developing countries are also affected. In the whole, this situation is to the contrary of the distribution of FF, which is reasonable considering the economic situation during this period. The following is some of the countries we selected:

 

References文獻

 

[1] Guidotti, P., Sturzenegger, F., Villar, A., On the Consequences of Sudden Stops  [J].Economía 4 (2), 1-44.2004.

[2] Kaminsky, G., and Reinhart, C., The Twin Crises: The Causes of Banking and Balance-of-Payments Problems [J]. American Economic Review 89 (3),473-500.2004.

[3] Kaminsky, G., Lyons, R. K., Schmukler, S., Managers, Investors, and Crises: Mutual Fund Strategies in Emerging Markets [J]. Journal of International Economics 64 113-134.2004.

[4] Kindleberger, C. P. With R. Aliber. Manias, Panics, and Crashes, A History of Financial Crises [J].Wiley, 2005.

[5] Kahneman, D. Tversky, A., Prospect Theory: An Analysis of Decision Under Risk [J]. Econometrica, 47 (2), 263-291.1979.