Empirical Results

A. Baseline Results

We started with the baseline regression (regression 1) and went on to add dummy interaction terms in order to differentiate between developing Asia and the OECD. The results are summarized in Table 4 [ PDF 25.7KB | 1 page ].

Our base-line regression (regression 1) suggests that larger countries received greater volumes of FDI and that the results are statistically significant. However, the coefficient of the source country is negative and economically and statistically significant. This result is not completely unexpected, as major source economies such as Japan and the US, and smaller source economies such as the Netherlands; Hong Kong, China; and Singapore are both major sources of FDI to the region. Possessing a common language is positively associated with increased FDI inflows, though it is not statistically significant. This may be because English is the de facto language of economic transactions in most of Asia. Greater distance between the host and source country appears to hinder bilateral FDI and this result is strongly significant, with the distance elasticity at about -0.4. Bilateral imports are positive and statistically significant. There is also evidence that a country that is more open to international trade and capital flows (based on the Chinn-Ito index) receives more FDI.

B. Robustness Checks

We undertook three robustness checks. First, given that annual data could be volatile, we reestimated regression 1 using three year data averages (regression 2). The results are quite close to the baseline, with the exception that financial openess in the host country loses its statistical signficance. Second, given the importance of the Hong Kong, China-PRC bilateral FDI flows, and the likelihood that a large part of that may be round-tripping, we re-estimated the regression by including a Hong Kong, China-PRC dummy (regression 3). Once again the results are robust with the exception being the lag of imports, which now loses its statistical signficance. Third, we re-estimated the regression by using OLS and converted the dependent variable to (1 + FDI) (regression 4). The results remain robust, with the exception of the common language dummy which becomes statistically significant. Overall, the results are highly robust.

C. Role of Time Zone Differences

Could the distance variable be capturing factors other than physical distance? In a recent paper, Stein and Daude (2006) emphasized the importance of time zone differences using OECD data for 17 OECD source economies and 58 host economies. According to the authors:

(t)he transaction costs associated with the difference in time zones should be important in activities that are intensive in information and require a great deal of interaction in real-time. Frequent real-time communications should be particularly important between headquarters and their foreign affiliates, as well as between a firm and its prospective foreign partners (p.97).

We therefore re-estimated regression 1 by including the difference in time zone between the host and source countries (regression 5 in Table 5 [ PDF 25.4KB | 1 page ]). The source data on time zone differences is from Stein and Daude (2006), and similar to their approach, we extracted the absolute difference of time between the host and source countries. Interestingly, with the inclusion of the time zone difference variable, the distance elasticity declines by about half in absolute terms to -0.2. The time zone difference variable almost wholly captures this decline in elasticity. In particular, the time zone elasticity is estimated at -0.18 and it is strongly statistically significant. All the other estimated coefficients are the same as in regression 1. It clearly appears that physical distance is partly affected by the differences in time zones between countries.

D. Intraregional versus Extraregional FDI Flows

While we do not have sectoral data on FDI flows, we recognize that there could be differences in the determinants of between FDI from the OECD and that from other developing Asian countries, particularly in consideration of distance and time zone variables. To determine this we re-estimated regression 1 by including an interaction term for all the dependent variables with the OECD economies (regression 6). Compared to the baseline regressions, the elasticities of host and source GDPs increases somewhat and remain statistically significant. Interestingly, although the elasticity of the source countries' GDP remain negative, the elasticity of the distance variable rises sharply to -0.65 and remains highly statistically significant, while the lag of imports becomes economically and statistically insignificant.

The OECD interaction terms offer some noteworthy findings, including the decline of elasticity of the host and source countries' GDPs (in absolute terms). Another notable finding is the statistical significance and sharp rise of elasticity of imports, which imply that bilateral imports tend to strongly attract bilateral FDI from the OECD economies, but not from developing Asian sources. This may suggest that FDI flows from OECD may be relatively export-oriented, while those from developing Asia are focused on the domestic market. This said, the distance elasticity of FDI from OECD sources sharply declines (-0.653+0.509=- 0.142), appearing quite counter-intuitive and requiring further examination. To this end we re-estimated regression 2 but only included the interaction terms for the non Asia-Pacific OECD economies (i.e., excluding Japan, Australia, and New Zealand). The results are outlined in regression 7. Interestingly, the non-interaction terms are quite similar to the baseline in regression 1, with the exception of financial openess which is no longer statistically significant. With regard to the interaction terms, it is notable that the distance elasticity is much higher (in absolute terms). Specifically, intraregional FDI is -0.4, while extraregional flows is a rather high -1.8.

Accordingly, we included the time zone difference variable and re-estimated the regression (regression 8). Notably, the time zone difference variable is no longer statistically significant, as expected a priori, since these are only intraregional flows. The interaction terms, economic masses, trade and financial openess, and common language elasticities, are all statistically insignificant, suggesting no difference between non Asia-Pacific OECD FDI sources and intraregional sources. Interestingly, the distance variable is also statistically insignificant⁹ The time zone difference is statistically and economically significant (entering with a negative coefficient), suggesting that there is no obvious difference in distance elasticity between intraregional and extraregional FDI flows to developing Asia. However, there is clearly a time zone difference effect that, ceteris paribus, reduces the amount of FDI flows to developing Asia from extraregional sources. One other interesting discovery is the significant rise in import elasticity (0.167+0.479=0.666). This suggests that extra-regional FDI is much more sensitive to bilateral import flows than intraregional FDI, and that FDI from extra-regional sources uses developing Asia relatively more intensively as a source of imports (i.e., as an export platform back to the home country). This is also consistent with the fact that many of the developing Asian economies run bilateral trade deficits with the US and EU.

Download this Paper [ PDF 110.6KB| 17 pages ].

Post a Comment

We welcome your feedback on this publication. Post a comment. ADBI is not obliged to acknowledge or publish comments and may abridge or edit them before web posting.

Comment(s)

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