Korean Migration to North America : Some Prices That Matter

The empirical migration literature has emphasized the role that differences in the return to human capital play in the migration decision. In this paper, we argue that many migrants are also concerned with differences in the return to the financial capital that they bring with them. One testable implication of the theory is that depreciation in the value of the Canadian dollar relative to the U.S. dollar should cause some migrants to substitute Canada for the United States as their destination of choice. Using data on Korean immigration to Canada and the United States, we estimate a regression model to test this hypothesis. The statistical evidence strongly supports a conclusion that exchange rate movements can cause some migrants to substitute destinations.


INTRODUCTION
Koreans have been immigrating to North America for just over a century, with the United States being the largest recipient of this migration flow. 1 As might be expected, the rapid convergence of Korean incomes with North America incomes has coincided with a decline in the flow of Korean migrants to North America.Despite this decline, Korean migration to Canada has recently grown (figure 1.1).Something has caused Korean migrants to substitute destinations.
Most economic theories of migration begin with the idea that migration flows are the result of decisions over where to invest human capital.A potential migrant evaluates the present value of their human capital in the home and destination country.If the present value in the destination country exceeds the full moving costs and present value of their human capital in the home country, then the potential migrant chooses migration.This human capital framework presents straightforward and testable implications for the pattern of migration flows.The most obvious are that: migrants tend to be young; migrants move from areas where the return to human capital is low to areas where the return to human capital is high; reductions in the costs associated with migration will increase migration flows; greater physical and cultural distances increase the costs and hence reduce the rate of migration.Not surprisingly, the evidence from a large empirical literature strongly supports these predictions (Ghatak et.al. 1996).
Further developments of the human capital story have led to other predictions.Todaro (1969) and Harris and Todaro (1970) introduced uncertainty into the migration decision by considering the impact of high unemployment in the migrant's destination.As with any other investment, the decision-maker would simply evaluate the expected net present value of the migration investment.The obvious prediction that increasing unemployment in the destination country reduces the expected return to human capital in the destination and so dampens migration flows has also been supported in numerous empirical studies.Other elaborations on the human capital story offer no easy solution to the puzzle.A more sophisticated model of heterogeneous labor that self selects was introduced by Sjaastad (1962), and developed by Borjas (1987, 1989, 1994) and Chiswick (2000).The usual predictions of the human capital story carry through, along with some unexpected predictions about how increasing inequality in the destination country increases immigration of individuals with large human capital endowments.Hatton (2003) uses the framework to show that an increase in the inequality in the destination country should increase the rate of flow of migrants.The puzzle this time is that while inequality rose in the Canada, it rose more quickly in the United States (figure 1.2).Again, the expectation created by theory is that Korean migration to Canada should be declining not rising.1 9 6 6 1 9 6 9 1 9 7 2 1 9 7 5 1 9 7 8 1 9 8 1 1 9 8 4 1 9 8 7 1 9 9 0 1 9 9 3 1 9 9 6 1 9 9 9 Canada Korea U.S.A.
Source: University of Texas Inequality Project (UTIP) data.
The 'new economics of migration' literature (Stark, 1991) has explored whether migration might usefully be examined as part of a portfolio investment decision of risk diversifying families.Accordingly, the expectation is that families would choose to invest their human capital assets over geographically dispersed and independent markets.Given the integration of Canadian and U.S. markets, it seems likely that risk diversifying family investors would have evaluated a Canadian situated human capital asset as being outside of the efficient portfolio.Again the human capital theory prediction is at odds with experience.
Other human capital models have emphasized search costs (Maier, 1985 andBerninghaus andSeifert-Vogt, 1991), the option value of waiting (Burda, 1995), liquidity constraints (Massey, 1988, Ghatak and Levine, 1994, and Hatton and Williamson, 2002), and social network effects (Bauer and Zimmerman, 1995).None of these provides any obvious explanation of the rapid increase in Korean migration to Canada in the 1990's.When the human capital model is extended to examine the impact of non-price factors such as political stability, political rights, and individual freedoms in the source country, the consistent finding is that they are all important determinants (Kamemera et. al, 2000).Since Korean civil rights and political stability were increasing over this period, it is hard to understand why such changes would lead Korean migrants to substitute Canada for the United States as a destination.
To get a handle on the apparent linkages between Canada and the United States, it is necessary to develop a model that can detect substitution.Baker and Benjamin's (1995) examination of migration from the Asia-Pacific region to Canada estimated a number of human capital models consistent with the direction of most of the empirical literature, and found the usual strong evidence supporting the predictions of the Harris-Todaro model and the common prediction that population density in the source country is positively related to the rate of migration to a Canadian destination.When they extended their model to allow migrants to substitute between Canada and the United States, the model no longer performed well at all, unless country dummy variables were added to the model.Once, these country fixed effects were added, the expected substitution effect reappeared.Canadian convergence to U.S. capita GDP caused some immigrants to substitute Canada for the United States.The statistical importance of the country fixed effects and the fragility of their substitution finding to a small change in the specification caused them to conclude that "non-economic supply-side variables are the most important determinants of immigrant flows between the Asia Pacific countries and the United States and Canada" (318).
If we accept their conclusion, we should look to non-price social changes in Korea to explain the massive substitutions occurring over the second half of the 1990's.One obvious possibility is that the move to democracy has paradoxically increased friction with the United States.At the same time, the number of Canadians working in Korea as teachers has grown dramatically to the point where Canadians have become the largest group of foreign teachers working in Korea.The combination of potent anti-Americanism and personal connections to Canada through teachers might have made Canada appear more attractive to prospective migrants.
An alternative possibility is that U.S. and Canadian demand for immigrants is the determining factor in the measured flows.DeVoretz (1995,(349)(350)(351) points to evidence in Green and Green (1995) to suggest that Canadian tinkering in immigrant admissions criteria over their 1974-92 study period prevented Baker and Benjamin from identifying the supply relationship that they were attempting to estimate.Interestingly, Green and Green (1999) have recently presented evidence showing that in the late 1980's the Canadian government abandoned its attempts to match inflows with domestic labor market conditions, "switching almost completely to long-term goals" (1999,447).With this switch to a policy of steady demand for immigrants, the identification problem recedes.
Given that the rapid growth in Korean migration to Canada occurred during a period of stable immigration policies, attention to changes affecting the supply of Koreans to North America seems reasonable.Interestingly, the rapid growth in Korean migrants choosing Canada

MODEL
There is evidence in the regional migration literature suggesting that physical capital and its price play an important role in the migration decision.In Lucas ' (1985) study of rural-urban migration in Botswana, a multinomial logit model confirmed the predictions of the Harris-Todaro model and indicated that the number of cattle owned played a role in the decision to migrate.In developed country studies of internal migration, differences in housing costs were shown to strongly influence regional migration in Italy (Attanasio and Padoa Shioppa, 1991), in Spain (Bentolila and Dolado, 1991), and the United States (Gabrel, Shack-Marquez and Wascher, 1992).The implication is that households bring more than human capital with them.
To introduce this into an international migration setting, consider the following model of a Korean migrant's decision to move to North America.As with most human capital stories, each family (i) is endowed with H i units of human capital which (with an appropriate choice of units) yield an expected present value of H i $U.S.In the destination country, this human capital might be expected to earn a premium w c in Canada and a premium w us in the United States of America.The expected present value of the family's human capital therefore be w c * H i in Canada and w us * H i in the United States of America.All units are in U.S. dollars.
In addition, the family might also posses K i units of a capital good which (with an appropriate choice of units) has a liquidation value of P k K i won, where P k is the price per unit of capital.This could be converted into U.If each family's preferences can be represented by some utility function that is increasing in its two arguments, then it is clear that rate of Korean migration to Canada (to the U.S) will be increasing (decreasing) in w c /w us and increasing (decreasing) in (e c /e us )(P us /P c ).While it might be possible to specify a utility function and derive an estimation equation directly, we prefer to apply some of the empirical models in the literature to Korean migration to North America.
The first model we wish to investigate is the Baker and Benjamin (1995) model used to examine Asia Pacific immigration to Canada.Adopting this to the story of Korean migration to North America yields: Baker and Benjamin then differenced the two equations and assumed that β C1 = β U1 and β C4 = β U4 .This they argued was equivalent to assuming that "from the individual migrant's perspective, a change in local income or local population density had the same impact on their propensity to migrate to either the United States or Canada, holding U.S. and Canadian opportunities constant" (315).The result was the following estimating equation.
This empirical model implicitly assumes that the premium to human capital in Canada and the U.S. (w c /w us in the previous discussion) can be proxied by the relative difference in real GDP per capita and differences in the unemployment rate.2Our extension builds on this practical compromise by introducing the difference between the exchange rate in Canada and the United States as an added explanatory variable.The resulting model would is: where ∆ln(X t ) is the Canada-U.S. difference in the log of the exchange rate.The expectation is that depreciation in the Canadian currency relative to the U.S. currency should cause some Korean migrants to consider substituting Canada for the United States as a destination.We also expect that auto-correlated errors are likely to be a problem, and that the differencing might transform the data sufficiently to reduce the influence of auto-correlated errors on the estimation.Hatton (2003) provides an alternative approach by estimating the determinants of migration flows in a random effects panel model.Given the limited data we have at our disposal, it is difficult to adapt such a model to the task of examining the substitution effect.The model in equation 4 allows testing for a substitution effect in a parsimonious manner.

RESULTS
The point of departure for our study was the Baker and Benjamin (1994) paper and their results indicating that country fixed effects were particularly important.Our first step was to reexamine their two models using a longer time series.The results of the estimation of equations 1, 2, and 3 together with the Baker and Benjamin estimates are presented in Figure 3.1.There are a couple of remarkable differences.First of all, the impact of Korean GDP on migration flows seems to be unimportant.Second, the impact of Canadian GDP on Korean migration to Canada seems to be opposite to expectations.So also is the impact of U.S.
unemployment on Korean migration to the United States.Perhaps the most surprising result is that the population density variable appears to be significant in both the Canadian and the U.S.
regression, but the sign is significantly negative in the United States.3One likely explanation is that the steady increase in Korean population density simply acts like a time trend variable, furthering the suspicion of specification problems in the Baker and Benjamin model.
We explored many sensible extensions of the Baker and Benjamin model to see if the addition of plausible variables might deliver estimates that are more consistent with theory.By and large, the results were far from satisfactory.The most obvious extension was to consider whether unemployment in Korea was having a significant impact on the migration decision.
Presented in Figure 3.2 as equations 1.1 and 1.The low Durban-Watson statistic indicates that autocorrelation is a problem.Attempts to correct for this using a Cochrane-Orcutt type procedure did change the sign on the ln(PCGDP CANt ) variable but the R 2 dropped to about 0.5 and none of the variables in the regression were significant according to a standard t-test.
Rethinking the problem, we noted that Green and Green (1999) had identified a change in Canadian Immigration policy in the late 1980's. 4The government abandoned fine tuning the points system to match immigration flows to local labor market conditions.To capture the impact of these changes, we introduced a structural dummy variable for the years 1988 through to 2003.The dummy variable alone or interacting with the other variables did seem to improve the fit somewhat, but the model still showed the problems identified in the earlier specifications.
Specifically, the parameter on ln(PCGDP CANt ) was still negative, and the Durban-Watson statistic remained low.
The hypothesis that we have advanced in the previous section is that exchange rates may   The finding of a serious specification problem is consistent with our argument that changes in exchange rates may cause price substitution by migrating Koreans.Equations 1 and 2 are not well equipped to uncover this substitution.As discussed earlier, Baker and Benjamin's second model (equation 3) offers a more natural model for examining the impact of exchange rate movements in the migration decisions (equation 4).
Before estimating equation 4, we re-estimated equation 3 for Korean migration to North America.The results are suggestive as the estimates show an apparently strongly significantly negative parameter value for the difference in Canadian and U.S. income.This should not be that surprising as Baker and Benjamin found that the positive estimate quickly became negative with the addition of country dummy variables and a time trend.The results also show that the DW statistic does not support a conclusion of at serial correlation in equation 3. Given all of this, the evidence seems to point to Korean migrants perversely choosing Canada over the United States when Canadian incomes fall relative to the United States.

CONCLUSIONS
"differences in net economic advantages, chiefly differences in wages, are the main causes of migration" (Hicks, 1932, p.76)The new migration literature has not substantially altered our understanding of the role that prices play in the migration decision.Instead of Hicks' emphasis on wage differentials, the new literature emphasizes differences in the return to human capital.While this is more nuanced than the earlier view, the literature has continued to view the price of labor as the price that matters.In this paper, we have attempted to outline reasons why the price of currency matters.
While the results are preliminary, the evidence is consistent with a view that relatively prosperous migrants are influenced by exchange rate movements in deciding between potential destinations.
More generally, we have initiated an examination of migration flows in which migrants substitute destinations when prices change.The empirical and theoretical literature is virtually silent on this important aspect of economic choice.Our contention is that the relative importance of country dummy variables in the empirical literature may at least be partly an artifact of a focus on the determinants of national in-migration or out-migration. 5Such models cannot easily capture substitution of one destination over another by migrants.Unfortunately, our results are not yet strong enough to allow us to draw definitive conclusions about whether such substitutions are as important as we suspect.To explore this more thoroughly, we need to determine whether similar results can be found with other relatively prosperous migrant groups.We also need to think more carefully about how to empirically model migration within a demand theoretic framework.

Figure
Figure 1.1

Figure
Figure 1.2 as a destination has a parallel in the rapid growth in Korean visitors and Korean students choosing Canada as a destination (figure1.3).This parallel movement points toward some common cause.In the next section we propose a simple extension of the human capital model to consider how moving other capital assets abroad might influence the migration decision.The model suggests that the relative movements in exchange rates might have played an important role in causing Koreans to substitute Canada for the United States as a destination.The model is then estimated and conclusions drawn.

Figure 1. 3
Figure 1.3 play a role in the location decision of migrating Koreans.Unfortunately, extending the model to consider the impact of foreign exchange rates facing Koreans leads only to a marginal improvements.One interesting result is that the Canadian cost of U.S. dollars provides sharper estimates than the Korean Won cost of Canadian dollars or the Korean Won cost of U.S. dollars.This suggests to us that there is non-zero cross-price elasticity between the Korean immigration to the U.S. and Canada.Presented below in figure 3.3 are the results of estimates from one of these extensions to equation 1.The results are not strong.The only parameter that seems remotely significant is the time dummy (D).The interactive dummy variables for Canadian GDP (DUMC) and Korean GDP (DUMK) have the expected signs but they are not strong estimates.The log of the Canadian cost of U.S. dollars (LGCXR) was the best performing exchange rate from those tested.The estimated value suggests that increases in the value of this variable (depreciation in the currency) cause immigration by Koreans to Canada to decline.This is completely counter to expectations.Not much should be read into this result as the t-statistic is low.More importantly, the Durbin-Watson statistic is low (DW=1.21),indicating important problems remain with the model.

Figure
Figure 3.3

Figure 3. 5
Figure 3.5 us H i , e us (P k /P us )K i ) > U(H i , K i ) and U(w us H i , e us (P k /P us )K i ) > U(w c H i , e c (P k /P c )K i ), S. dollars to provide e us P k K i U.S. dollars of financial capital, where e us is the won cost of $1 U.S. With this financial capital, the family can buy e us (P k /P us )K i units of capital in the United States.Similarly, the same family can purchase e c (P k /P c )K i units of capital in Canada.If the household's preferences can be represented by a utility function (U) then the household will decide to migrate to the U.S. if: U(w

Figure 3.1
2, the results very clearly indicate that Korean unemployment is not the missing piece.While it does cause the estimates to appear to firm up, the estimates continue to suggest that an increase in Canadian income reduces Korean immigration.Moreover, if we include the population density variable, the estimated parameters in the model do not look that different from the model without Korean unemployment.Intercept ln(PCGDP Kt ) ln(PCGDP CANt ) URATE CANt URATE KORt ln(POPDEN Kt ) R 2 DW 3.3 attention is turned to Korean immigration to the United States (equation 2), the regression results seem more in line of existing studies.According to the estimates, an increase in U.S. GDP is related to increases in Korean immigration to the United States, while an increase in Korean GDP is related (weakly) to decreases.The only peculiarity is the indication that increases in U.S. unemployment is related to increases in Korean immigration to the United States.As with the Canadian case, extension of the model to consider the impact of Korean unemployment (equation 2.1 in figure3.4)does not provide a demonstrable improvement.More importantly, the apparent significance of the trending population density variable remains and the Durbin-Watson statistic is low.Extending the model with time dummies and exchange rates does not bring any improvement, and more importantly does not eliminate the problem of auto- replaced by a more global outlook.At this time Canada emerged as a close alternative (substitute) to the United States in the minds of many Koreans.When While this perverse choice might be explained by other sociological factors, there remains the possibility that other prices matter, and the absence of these prices constitutes a specification problem.If our argument in the last section is correct, then it is likely that the exchange rate may be important.To investigate this, we estimated equation 4 to see whether the addition of the Canadian dollar cost of US dollars altered the estimates.The results, presented below in figure3.6,do not provide as clear an adjudication of the issue as might be desired.The Durbin-Watson statistic and R squares improve greatly, but the sign of the income variable continues to suggest that Korean migrants make perverse choices.The exchange rate variable has the expected sign, but the t-statistic is low.