Empirical Analysis

Here, we examine the effects of foreign capital inflows on various economic variables, especially asset prices, using a VAR (Vector Auto-Regression) model. VAR models provide a useful methodology for investigating this issue. First, VAR models are data based, with a relatively small number of restrictions. This empirical framework is useful for documenting empirical facts. Second, the effects are expected to be inherently dynamic. For example, foreign capital inflows may affect different types of asset markets with different timing. VAR models are useful for inferring dynamic effects.

Let us assume that an economy is described by the following structural form equation:

G(L) y_t = e_t

where G(L) is a matrix polynomial in the lag operator L, y_t is an m×1 data vector, m is the number of variables in the model, and e_t denotes a vector of structural disturbances. Constant terms are dropped for notational simplicity. Assuming that structural disturbances are mutually uncorrelated, var(e_t) can be denoted by Λ, which is a diagonal matrix where diagonal elements are the variances of structural disturbances.

We estimate the following reduced form panel VAR with the individual fixed effect:

y_t = B(L) y_t + u_t ,

where B(L) is a matrix polynomial in the lag operator L, and var( u ⁱ_t ) = Σ .

There are several ways of recovering the parameters in the structural form equation from the estimated parameters in the reduced form equation. The identification schemes under consideration impose recursive zero restrictions on contemporaneous structural parameters by applying Cholesky decomposition to the reduced form residuals, Λ, as in Sims (1980).

Note that our statistical inference is not affected by the presence of non-stationary factors since we follow a Bayesian inference (see Sims 1988 and Sims and Uhlig 1991).⁵

4-2 Empirical Model

In the basic model, the data vector, y ⁱ_t , is {Y, P, R, CAP_OUT, CAP_IN, X} where Y is the log of a measure of output, P is the log of the measure of price level, R is the interest rate, CAP_OUT is capital outflows or portfolio outflows, CAP_IN is capital inflows or portfolio inflows (as a ratio to trend GDP), and X is the domestic variable under consideration.⁶ For X, we consider the following set of variables: the log of the KOSPI 200 index (KOSPI200), the log of the KOSPI index (KOSPI), the log of KOSDAQ index (KOSDAQ), the log of the wondollar exchange rate (ERUS), the log of the won-yen exchange rate (ERJ), the log of the won-euro exchange rate (ERE), the log of the nominal effective exchange rate (NEER), the log of the real effective exchange rate (REER), the log of apartment price (APT), the log of housing price (HOUSE), the log of foreign exchange reserves (FRES), the log of monetary base (MB), the log of M1 (M1), and the log of M2 (M2). We included CAP since they are the main variable of our interest. Y and P are included to control for the factors that can affect X, including asset prices.

The factors or variables affecting domestic variable X can be divided into three types. First, certain factors affect X mostly through changes in foreign capital inflows. For example, a policy change toward a more open foreign capital market would affect capital flows and then affect X. Second, certain factors affect a domestic variable X mostly through channels other than foreign capital inflows. For example, an increase in the price level (which may be the result of a monetary expansion) may increase domestic asset prices, but in this transmission, foreign capital inflows are not likely to play an important role. Third, there are certain factors that affect X not only through changes in foreign capital flows but also through other channels. For example, a change in the domestic economic condition induces foreign capital inflows and then affects the domestic variable X. But a change in the domestic economic condition also influences investments by domestic investors and thereby affects asset prices.

The first type of factor affects X mainly through the changes in capital inflows. Therefore, to analyze the effects of capital inflows, it is unnecessary to control for this type of factor in the model. However, the second types of factor should be controlled because there may be an omitted variable bias if an important factor is not included in the model. On the other hand, we also try to control some third type of factors. If we exclude this type of factor in the model, all the effects of this factor, including the effects through channels other than changes in capital inflows, may be captured as the effects of foreign portfolio inflows.

As a second type of factor, we control for the aggregate price level. The aggregate price level shows the nominal and monetary condition of the economy, which can also affect X, for example, asset prices. As a third type of factor, we control for the domestic interest rate and aggregate output. Aggregate output is the most important variable representing the domestic economic condition, which may affect X both through changes in foreign capital inflows and through other channels. A change in the interest rate may affect asset prices directly, and also affect capital inflows. On the other hand, it may not be necessary to control some second types of foreign factors because their indirect effects are already captured in the control variables. For example, a change in the U.S. real economic condition may affect the domestic economy through real economic linkages, not by changes in capital flows. But if a variable reflecting the domestic economic condition (Y in our model) is controlled, such indirect effects can be controlled at least to some extent. Finally, we also control for capital outflows since capital outflows and inflows are sometimes inter-related, and we would like to separate the effects of capital inflows only.

Regarding the ordering of the variables, all the control variables are assumed to be contemporaneously exogenous to capital inflows in order to take out all the inter-related effects from capital inflows shocks. On the other hand, capital inflows are assumed to be contemporaneously exogenous to X. This type of assumption is used by Kim, Kim, and Wang (2004), Froot, O'Connell, and Seasholes (2001), and Bekaert, Harvey, and Lumsdaine (2002). In order to make the assumption more reliable, the data is constructed as of the end of the period value. Consequently, capital inflows are a flow variable that represents the activities during the period while X represents the value at the end of the period. Therefore, the assumption that other variables such as capital inflows are contemporaneously exogenous to X is a reasonable one.⁷

Finally, we note that the ordering among Y, P, R, CAP_OUT does not matter when we examine the effects of shocks to capital inflows.⁸ Monthly data is used for the estimations. The estimation period is from January 1999 to September 2007. We exclude the period prior to 1999 since economic behavior before and after the Asian crisis may be considered different within the framework of our study. A constant term and three lags are assumed. As a measure of output, we use price level, the interest rate, industrial production, CPI, and the call rate. To construct capital inflows and outflows, we exclude FDI since its effect may be somewhat different from the effects of usual capital flows.

4-3 Results

Figure 17 [ PDF 15.4KB | 1 pages ] and Figure 18 [ PDF 15.5KB | 1 pages ] report the impulse responses, with 90% probability bands for the 2-year horizon, of each variable to capital inflows shocks and portfolio inflows shocks, respectively. The names of the responding variables are reported at the top of each graph.

First, to discuss the nature of capital inflows or portfolio inflows shocks, we first examine the impulse responses of capital inflows or portfolio inflows. Typical capital inflows shocks involve an approximate 4% (as a ratio to trend GDP) immediate increase in capital inflows while a typical portfolio inflow shock involves an about 2.5% (as a ratio to trend GDP) immediate increase in capital flows. In both cases, the responses return the flows to the initial level very quickly, but the responses of portfolio inflows are a bit more persistent.

Both types of capital inflow shocks increase stock prices sharply on impact, but the effects of portfolio inflows are larger and more persistent. Capital inflow shocks increase the KOSPI index about 2% on impact while portfolio inflows increase it by about 3%. The KOSPI index returns to the initial level about four months after the capital inflow shock, and returns to the initial level about 1 year after a portfolio inflows shock. The effects on the KOSDAQ index are also large and significant. Capital inflow and portfolio inflow shocks, on impact, increase the KOSDAQ index by about 2.5% and 4%, respectively. The effect of portfolio inflow shocks is more persistent than that of capital inflows shocks.

On the other hand, the effect on housing and apartment prices are moderate and insignificant. The point estimate shows that the size of the change is relatively small, far below 5%. In addition, the 90% probability bands include zero responses in all cases. These small effects may be related to recent government policy measures for regulating the housing market in Korea.

The nominal and real effective exchange rates tend to appreciate in the very short-run. In the case of capital inflows, the impact effects of the approximate 0.25% appreciation are marginally significant. However, in the case of portfolio inflows, the probability bands are very wide, including zero responses. In both cases, the effects on the won-dollar exchange rate are also very small and insignificant.

The small effect on exchange rates seems to be mostly related to the foreign exchange intervention to accumulate foreign exchange reserves. In both cases, foreign exchange reserves increase significantly. In response to capital inflows shocks, foreign exchange reserves increase about 0.35% on impact and then increase up to about 0.5% within two or three months after the shock. In response to portfolio inflows shocks, foreign exchange reserves increase about 0.2% on impact and then increase to about 0.5% three months after the shock. As a result, the monetary aggregates like monetary base, M1, and M2 do not increase significantly.

4-4 Determinants of Capital Inflows

Here, we will briefly examine the determinants of capital inflows by modifying the empirical model. To evaluate the role of various factors, we include various factors explicitly in the model. In the previous model, four pull factors (domestic interest rate, price level, domestic output, stock price) were explicitly included. We also included two important push factors, world interest rate and world output. On the other hand, we excluded capital outflows; capital outflows were included in the model to isolate the effects of capital outflows shocks, and now we exclude it to preserve the degree of freedom.⁹ As a result, we construct a model of {Y^*, R^*, Y, P, R, CAP_IN, SP}, where Y^* and R^* are world output and world interest rate, respectively, and we order the contemporaneously exogenous ones first.

In the model, we assume that world variables are contemporaneously exogenous to Korean variables since Korean economy can be regarded as a small open economy that cannot affect world variables much. We also assume that output and the price level are contemporaneously exogenous to the interest rate since aggregate variables tend to move sluggishly but financial variables tend to respond to information instantaneously.^10,11 In the estimation, we use the U.S. variables as proxies for world variables.

To discuss the role of each factor, we report the variance decomposition of capital inflows and portfolio inflows in Table 2 [ PDF 17KB | 2 pages ]. From the result, one dominant factor does not emerge; each shock plays some role (about 5-10%) in explaining capital and portfolio inflow fluctuations. The role of two push factors is not very large; about 10% of capital and portfolio inflows fluctuations are explained by the two push factors. The role of each pull factor tends to be larger than that of each push factor. For capital inflows fluctuations, output shocks explain about 10%. For portfolio inflows fluctuations, stock price shocks explain about 10%.

Download this Discussion Paper [ PDF 443.4KB| 35 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.