RDP 2009-09: Volatility in International Capital Movements 4. Sources of Capital Account Volatility

This section attempts to explore some of the underlying sources of capital account variability using panel data regressions. Broadly speaking there are two types of shocks that could affect the capital account flows of a particular country. First, investors (for whatever reason) may change their views about the prospects for that country as a whole, leading capital to either flow into or out of that country in a way that implies a positive correlation of the components of the capital account. Second, a shock may be more specific to a particular asset class within a country (or across countries), but the extent to which this affects the overall volatility of the capital account and the correlations between its various components will depend on the degree of substitutability between different asset classes within that country.

We examine the extent to which factors that may affect the degree of substitutability within each country play a role in explaining volatility of the overall capital account. Of course, the results will only be suggestive of any substitutability effect, since changes in the nature of the country-specific shocks may also affect overall capital account volatility. We leave a more detailed investigation of this issue to future research.

4.1 Selection of Explanatory Variables and the Model

The dependent variable to be explained is the volatility of the overall capital account balance (as a share of GDP). Data are annual, and so we measure volatility as the standard deviation of quarterly capital account observations within a given year.

The first two explanatory variables we consider correspond to those discussed in Section 2, which reflect the importance of foreign direct investment (FDIshare) and bank and money market flows (BMMshare) in the capital account. We measure the importance of a (net) flow as the ratio of its absolute value to the sum of the absolute value of all flows (within each year).

We also control for the exchange rate regime. If the exchange rate is fixed or pegged, the burden of any external adjustment following a shock must fall more on quantities rather than prices, which may imply a more volatile capital account. To test for this we include a dummy variable that is ‘zero’ when the exchange rate is freely floating or under a managed float, and ‘one’ if the exchange rate regime is less flexible (FXregime); see Appendix B for further details.

We include two variables to capture the potential for a substitutability effect. Both are intended to capture the extent of the development or depth of financial markets; of course, this in turn is likely to reflect some deeper structural features of the various economies. The first variable is a measure of the volume of gross flows. Large two-way flows may limit volatility by enhancing the scope of residents to meet financing requirements at times when foreigners reverse their investments and become sellers.[11] The extent of gross flows may also reflect the degree of market development. For example, it would be unrealistic to expect portfolio debt flows to play a major part in smoothing capital flows in economies that do not have well-developed bond markets.[12] Furthermore, the extent of gross capital flows reflects the degree of capital account openness. The less open the capital account, the less scope there is for shocks to one type of flow to be offset by changes in other flows.

To gauge the importance of these factors, we construct a summary measure based on gross flows (FlowOpenness). We create an index that depends on the ratio of the absolute value of the gross flows to the absolute value of the sum of gross and net flows as follows:

When capital flows freely in both directions, we expect the sum of absolute gross flows to be large relative to net flows. In this case, the index tends towards 100.

When capital flows are very one-sided, we expect gross flows to be smaller relative to net flows. The most extreme case would be where gross flows are the same size as net flows. This would occur if resident or non-resident flows were completely restricted. In this case, the value of the index would be zero. To illustrate, the United Kingdom has an average openness index score of 90 over the period 1980 to 2005, the highest for any economy in our sample reflecting London's role as a global financial centre, while Thailand has an average score of only 8, the lowest average score of the countries sampled (see also Appendix C). In the panel analysis, the FlowOpenness variable is included contemporaneously and with a lag.

The second variable intended to capture any substitutability effect is a measure of the degree of domestic financial market development. This is likely to be an important determinant of the ability of investors to substitute between different forms of finance. We use the ratio of equity-market turnover to market capitalisation as a proxy for financial market development (MarketDevelopment).

In addition to these variables, we allow for unobserved time-invariant factors to influence the volatility of each economy's capital account by using a fixed-effects estimator.

In summary, the regression we estimate is of the form:

where: Inline Equation represents the volatility of the capital account as a ratio to GDP for country i in year t; ηi is the fixed effect for country i; and εit is the error term.

We use a balanced panel of annual data for our 12 countries over the period 1991 to 2005.[13]

4.2 Regression Results

The results in Table 3 suggest that the composition of the capital account is not a significant determinant of its volatility – there is no statistically significant relationship between the share of foreign direct investment or bank and money market flows and the volatility of the total capital account. The coefficients on market development and financial openness (lagged) are negative and statistically significant (at the 5 and 10 per cent levels, respectively). This is tentative evidence in support of the idea that more developed financial markets might encourage greater substitutability between different types of flows, thereby helping to reduce the volatility of the overall capital account.

Table 3: Panel Data Estimation Results
Dependent variable is the volatility of the capital account to GDP
Coefficient value P-value
Constant 1.97 0.00
FDIshareit −1.07 0.24
BMMshareit −0.20 0.75
FlowOpennessit −1.4 x 10−3 0.33
FlowOpennessit-1 −3.5 x 10−3 0.06
MarketDevelopmentit −0.46 0.04
FXregime 0.15 0.44
R2 0.64  
Number of observations 180  
Wooldridge test for autocorrelation 0.00  

The coefficient on the exchange rate regime dummy was found to be statistically insignificant. However, care should also be taken when interpreting this result. In particular, we caution against interpreting this as evidence that fixing the exchange rate will not affect capital account volatility. Given that the capital account, current account and the exchange rate are jointly determined, it seems probable that fixing one of these variables would have some effect on the others. Indeed, when we run the above regression using a sub-sample of just the industrialised countries, the coefficient on the exchange rate variable is positive and statistically significant.


There is some evidence to suggest that industrialised economies are well placed to benefit from large overall gross flows, with both residents and non-residents playing sizeable roles in the external accounts. In contrast, emerging economies are characterised by being somewhat bank dependent and typically have gross flows dominated by the actions of foreigners (Lowe 2009). A history of large two-way flows may indicate an increased ability for residents to offset volatility caused by foreign investors, in part because it may be associated with residents accumulating a larger stock of foreign assets. [11]

Refer also to Debelle and Galati (2005). [12]

There may be an issue of endogeneity, whereby some of the right-hand-side variables are influenced by the same shocks that affect the overall volatility of the capital account. We leave this issue for future research and note for now that the results here need to be interpreted with this caveat in mind. [13]