Introduction
Among the principal analytic payoffs of systematic, cross-national research in the field of immigration studies is that of setting limits on generalizations derived from single-country studies. In the case of the United States, the most commonly used model of wage determination among immigrants treats their earnings largely as a function of standard human capital variables, especially the number of years of education upon arrival in the U.S. (see, for example, Borjas 1995), sometimes with a bow toward aggregate labor supply and public policy variables (Bansak and Raphael 1998; Dávila and Pagán 1998; Phillips and Massey, forthcoming). As part of a comparative study of the role of immigrant labor in the Japanese and U.S. economies (Cornelius and Kuwahara 1998), we have estimated multivariate models of wage determination among samples of foreign workers interviewed during the first half of 1996 in the Japanese industrial city of Hamamatsu and San Diego County, California. In addition to the usual human capital variables, these models incorporate a number of variables that cannot be measured using census data (e.g., decennial census data and the U.S. Current Population Survey), such as legal status, mode of job-seeking, and characteristics of the firm in which the immigrant is employed. The result is a much fuller picture of the wage determination process among foreign workers in the United States and Japan, which also reveals important variations in the explanatory power of certain individual characteristics from one national context to the other.

The dependent variable in both models is the natural log of hourly earnings, which enables us to interpret the estimated coefficients as the impact which the independent variable has on earnings as a percentage. Tables 1 and 3 present the means and standard deviations of the variables used in each model, in addition to the expected signs of the estimated coefficients. The model for foreign workers in San Diego incorporates some data that are not available for our Hamamatsu sample. Specifically, it is not possible in the case of Hamamatsu workers to match characteristics of the firm in which they are employed with the workers, except for the broad sector of the economy to which the firm belongs (manufacturing, construction, or services). The sampling procedure used for the San Diego portion of the study generated detailed data on the firm in which the immigrant was working, including the percentage of foreign-born persons in the production work force, the annual turnover rate among production workers, and whether native-born U.S. workers are part of the firm's applicant pool for entry-level production jobs. The better performance of the San Diego wage determination model (which explains 50 percent of the observed variation in wages) as compared with the Hamamatsu model (which explains 33 percent of the variance) can be attributed largely to the availability of additional information on the San Diego workers, which enabled us to include more of the theoretically relevant explanatory variables.

The San Diego model includes nine industrial-sector dummy variables, representing the nine sectors of the San Diego economy in which immigrant workers and their employers were sampled. The reference sector is the restaurant sector, in which average wages for entry-level workers were reported (by both employers and workers) to be the lowest; thus the signs of the estimated coefficients for being employed in the other eight sectors represented in the model are anticipated to be positive. In the Hamamatsu model the reference industry sector is the relatively high-paying service sector; hence the expected signs for employment in the manufacturing and construction sectors are negative.

Results
In our model of wage determination for Hamamatsu foreign workers (Table 2), only six of the independent variables proved statistically significant at the p<.05 or p<.10 level: gender (female), time lived in Japan (months, cubed), whether the worker found his or her current job through a relative or friend (i.e., social network-mediated job seeking), whether the worker found his current job through a professional labor broker, whether the worker was a Nikkeijin (foreign worker of Japanese ancestry from Brazil or Peru, admitted to Japan under a special visa category), and whether the worker was employed in the manufacturing sector.

As expected, given the traditionally high degree of wage discrimination against women in Japan, being female is strongly and negatively related to wage levels. Job recruitment modes (either being hired via a social network contact, or through a labor broker) are strongly and positively related to wages, as is being a Nikkeijin. As previously noted, the negative coefficients for being employed in construction or manufacturing mean that workers in those sectors are learning less than those employed in services (e.g., manufacturing workers earn 27 percent less per hour; construction workers earn 26 percent less, but that result is not statistically significant).

The variables that fail to be significant predictors of foreign worker wages in Hamamatsu are of particular interest. They include all of the socio-demographic and labor market experience variables that might be expected to boost wages: age, education, Japanese language ability, time spent in Japan (only one of the three forms of this variable is statistically significant, and barely so), and job seniority (years with the current employer). Similarly, illegal immigration status has no predictive power. In short, what foreign workers bring with them to Japan in terms of human capital (or which they may acquire after arrival, e.g., Japanese language competence) has far less influence on the wages they earn there than the auspices under which they gain employment. Only two personal attributes are important predictors: gender and nationality/ethnicity (Nikkeijin status).

The pattern of results from our regression model analyzing immigrants' earnings in San Diego County (Table 4) is quite different. In San Diego, all of the standard human capital variables are statistically significant predictors (especially English language fluency), as is job seniority (years with current employer). However, there are also some similarities in the results. Legal status (entered illegally) is not a significant predictor of wages, as in Hamamatsu; getting a job through a relative or friend is also significantly related, as in Hamamatsu, but in the opposite direction: In San Diego, if an immigrant got hired via relatives or friends, his wages were 8 percent lower than if they were not recruited through a social network. As expected, the higher the percentage of foreigners in the production work force at the immigrant's place of employment, the lower the hourly wages received; this is consistent with our previous finding that the most immigrant-dependent firms in San Diego pay lower wages, averaging $5.04 per hour for entry-level production workers as compared with $6.02 per hour in less immigrant-dependent firms (Cornelius and Kuwahara 1998: 20). We found this wage differential to be much less pronounced in Hamamatsu, where highly immigrant-dependent firms were paying only $0.12 per hour less for entry-level production work than less-dependent firms (not a statistically significant difference). Also, the model for San Diego shows that if some native-born U.S. workers apply for the same kinds of jobs done by immigrants in a given firm, an immigrant employed there is likely to be earning 10 percent more than his counterparts in firms where native workers are absent from the job applicant pool. This is a clear indication of upward pressure on wage scales, in firms that are not completely dependent on immigrants as production workers. Finally, all of the dummy variables in the San Diego model for sector of employment are statistically significant predictors of wages. For example, immigrants employed as construction workers earn 49 percent more than those working in restaurants.

Discussion
One of the most striking differences in the wage determination estimates for Hamamatsu and San Diego concerns the explanatory power of standard human capital variables. The absence of a statistically significant relationship between these variables and wages among foreign workers in Hamamatsu is quite consistent with what in-depth, unstructured interviews with employers and Brazilian Nikkeijin workers in various Japanese cities have revealed (see Tsuda 1998; forthcoming). Take, for example, educational attainment: Japanese employers and professional labor brokers typically don't ask foreign workers about educational level when hiring, because they know that a college or more advanced degree earned abroad has no relevance to the kind of unskilled work that foreign workers do in Japan. This includes the Latin American Nikkeijin, who are often well-educated, white-collar workers in their home country. Some Japanese employers even claim that the better-educated Nikkeijin perform worse as unskilled workers because they have been spoiled by air-conditioned offices and other privileged working conditions for so long that they are not sufficiently prepared to cope with the physical rigors of factory work.

Having Japanese language competence is considerably more important than years of education in getting a job; foreign workers who speak Japanese well are more likely to be hired by both brokers and direct employers. But, as our regression analysis reveals, Japanese language competence generally does not influence wages. The only observable subgroup of foreign workers who are rewarded for their Japanese proficiency are those who are promoted to lower-level supervisory jobs or who become Nikkeijin cultural intermediaries in company or broker offices, and there are very few such foreign workers in Japan.

Similarly, time spent in Japan and years of service to a given employer usually are not reflected in higher wages, because even the foreign workers who have lived in Japan longer continue to do unskilled factory work and do not advance to more skilled and higher-paying jobs within the company. This is especially true for foreign workers who are recruited by labor broker firms, which have traditionally supplied Japanese employers with workers to fill casual, disposable jobs. This casual work force previously consisted of native-born Japanese female part-timers and seasonal internal migrants (dekasegi workers), the supply of which eventually dried up, creating labor shortages that were met by foreign workers, beginning in the late 1980s.

"Brokered" workers are hardly ever hired permanently or promoted within the company; they are strictly "borrowed" by employers from their labor brokers and get "returned" to the brokers whenever production slackens. Foreign workers who are hired directly by Japanese employers are in a somewhat different situation. Some of them are promoted to a limited extent if they remain with the company for a very long time. However, like "brokered" workers, very few of the directly hired are ever made "seishain" (regular, "permanent" employees) and put on the promotion track like ordinary Japanese workers, receiving gradual salary increases. They remain temporary, marginal workers whose short-term contracts may not be renewed. Moreover, Nikkeijin workers tend to switch jobs so frequently that few stay around long enough in one company to move up to more skilled or higher-status jobs. Both factors help to explain why neither age, nor length of stay in Japan, nor years with the foreign worker's current employer leads to higher wages. Both older and younger foreign workers are hired for similar, temporary, unskilled jobs at the entry level, and most do not stay in their jobs long enough for age to start making any difference.

Human capital and labor-market experience variables have a much greater impact on immigrants' wages in San Diego County because, in a more mature and extensive immigrant labor market like that of San Diego, there is significantly greater variation in job skill level, and a much higher proportion of the jobs available to immigrants are permanent or at least long-term in character (see Cornelius 1998). As a consequence, there are numerous immigrant workers who stay in the same company and move up the job ladder, which has a significant influence on wages over time. This is reflected in relatively low labor turnover rates in the immigrant-dependent firms that we studied in San Diego County. In fact, the most highly immigrant-dependent firms have lower turnover among production workers (13 percent annually) than less immigrant-dependent firms (17 percent). In the case of Japanese firms, it is unlikely that a "jobs ladder" for foreign workers will begin to emerge as long as the labor brokerage system persists. That system strongly encourages the casual use of foreign labor, and all indications are that the involvement of brokers in recruitment of foreign workers in Japan is increasing rather than diminishing (Iguchi 1998).

Another intriguing difference between the results from Hamamatsu and San Diego relates to the wage effects of job-seeking through migrants' social networks. In Hamamatsu, being hired through referrals by relatives or friends raises wages by 20 percent, while in San Diego, obtaining a job through one's social network lowers wages by about 8 percent. Why should networks generate "negative social capital" (Portes 1998: 15-18) for immigrant workers in San Diego, while leading Hamamatsu's foreign workers to higher-paying jobs? In Hamamatsu, the Nikkeijin use their social networks not just to find initial employment but to secure a succession of higher-paying jobs. They constantly exchange and compare information about the wages and amounts of overtime offered by brokers and employers, and they quickly switch to their relatives' or friends' brokers or companies if compensation is higher there. Foreign workers who have less extensive social networks in Japan have fewer opportunities to find better-paying jobs and tend to be stuck with lower wages.

Immigrant workers in San Diego seem less intensely preoccupied with pursuing higher wages through inter-firm mobility. The proportion of permanent settlers is much higher among foreign workers in San Diego than in Hamamatsu (63.9 percent vs. 9 percent); the Hamamatsu foreign worker population consists to a far greater extent of those with a strictly instrumental, sojourner mentality ("I want to earn as much money as possible in a short period of time"). Accordingly, migrants' social networks in San Diego perform a broader range of functions, beyond simply providing access to better-paying jobs. They track newly-arriving migrants into companies employing clusters of relatives and friends, where immigrants constitute a large majority of all production workers, where Spanish is the predominant language of the work place, and where firm owners are more likely to be immigrants themselves. Immigrant-owned businesses in San Diego have almost exclusively immigrant labor forces and tend to pay lower wages, holding constant other relevant firm characteristics. Yet labor turnover rates in immigrant-owned businesses in San Diego are lower than in U.S. native-owned firms, which suggests the importance of a workplace culture in which ethnic solidarity, ease of communication (in Spanish), and family ties provide incentives for stability. Thus, immigrant workers in San Diego are compensated, in other ways, for the relatively small wage penalty that they pay for obtaining jobs through their social networks.

To summarize: The generally uniform, unskilled nature of jobs held by foreign workers in Japanese cities like Hamamatsu, the workers' tendency to stay on the job for only short periods (especially true of the Latin American Nikkeijin), and their reliance on labor brokers effectively confine them to the peripheral sector of the Japanese labor market and inhibits their long-term economic mobility (in wage and occupational status terms). In the case of San Diego County, we find a larger and more diversified immigrant labor market, with a greater variation in job types, ranks, skill levels, and job stability; very little labor-brokerage activity (even in the agricultural sector); and the presence of more long-staying migrants or permanent settlers who have been able to escape the peripheral market of casual jobs and move up to higher status, more stable, and better-paying employment, often by acquiring additional human capital (especially English language competence) that is rewarded by employers.

It remains to be seen whether labor markets for foreign workers in Japanese cities like Hamamatsu will evolve toward the San Diego model, as permanent settlement inevitably becomes more common (Tsuda 1998; forthcoming), or whether Japan's elaborate labor brokerage system is so deeply institutionalized that foreign-worker wages will continue to be determined more by labor recruitment mechanisms than by individual human capital endowments and labor-market experience.

REFERENCES CITED

Bansak, Cynthia, and Steven Raphael (1998) "Immigration Reform and the Earnings of Latino Workers: Do Employer Sanctions Cause Discrimination?" Department of Economics, University of California-San Diego, Discussion Paper 98-20 (August).

Borjas, George J. (1995) "Assimilation and Changes in Cohort Quality Revisited: What Happened to Immigrant Earnings in the 1980s?" Journal of Labor Economics 13 (2): 201-245.

Cornelius, Wayne A. (1998) "The Structural Embeddedness of Demand for Mexican Immigrant Labor: New Evidence from California," pp. 114-144 in Marcelo Suárez-Orozco, ed., Crossings: Mexican Immigration in Interdisciplinary Perspectives. Cambridge, Mass.: Harvard University Press.

Cornelius, Wayne A., and Yasuo Kuwahara (1998) The Role of Immigrant Labor in the U.S. and Japanese Economies: A Comparative Study of San Diego and Hamamatsu, Japan. La Jolla, Calif.: Center for U.S.-Mexican Studies, University of California-San Diego (April).

Dávila, Alberto, and José A. Pagán (1998) "The Impact of IRCA on the Job Opportunities and Earnings of Mexican-American and Hispanic-American Workers," International Migration Review 32 (1): 79-95.

Iguchi, Yasushi (1998) "International Migration in East Asia: A Growing Challenge for Japan," paper presented at the Fall Workshop of the University of California Comparative Immigration and Integration Program, University of California-Davis, October 9-10.

Phillips, Julie A., and Douglas S. Massey (forthcoming) "The New Labor Market: Immigrants and Wages After IRCA," Demography.

Portes, Alejandro (1998) "Social Capital: Its Origins and Applications in Modern Sociology," Annual Review of Sociology 24: 1-24.

Tsuda, Takeyuki (1997) "Ethnic Transnationalism and the Channeling of Migrant Flows: The Return Migration of the Japanese-Brazilians," pp. 35-63 in Diane Baxter and Ruth Krulfeld, eds., Beyond Boundaries: Selected Papers on Refugees and Immigrants, Vol. 5. Arlington, Va.: American Anthropoloigical Association.

------- (1998) "From Ethnic Affinity to Alienation in the Global Ecumene: The Cultural Encounter Between the Japanese and Japanese-Brazilian Return Migrants," paper presented at the conference on "Cultural Encounters Between Latin America and the Pacific Rim," University of California-San Diego, March 6-7.

------- (forthcoming) Strangers in the Ethnic Homeland: The Return Migration of the Japanese-Brazilians and the Making of Japan's New Immigrant Minority. Chicago: University of Chicago Press.

Waldinger, Roger (1997) "Social Capital or Social Closure?--Immigrant Networks in the Labor Market." Lewis Center for Regional Policy Studies, University of California-Los Angeles, Working Paper No. 6.

Table 1: Definition and Characteristics of Variables Used in
Hamamatsu Wage Determination Model

Variable Measurement Expected Mean Std. Dev.
Sign

Female (dummy) Female = 1, Male = 0 - 0.429 0.496
Age Years + 30.313 8.649
Years of schooling Years + 12.719 3.585
Speaks/understands Japanese well (dummy) Yes = 1, No = 0 + 0.415 0.494
Time living in Japan Months + 46.475 31.062
Entered Japan illegally (dummy) Illegal = 1, Legal = 0 - 0.065 0.246
Found job through friend/relative (dummy) Yes = 1, No = 0 +,- 0.410 0.493
Work through labor broker (dummy) Yes = 1, No = 0 +,- 0.350 0.478
Years with current employer Years + 1.871 1.754
Nikkeijin (foreigner of Japanese ancestry) Nikkeijin = 1, non-Nikkeijin = 0 + 0.751 0.433
Construction sector (dummy) Yes = 1, No = 0 - 0.023 0.150
Manufacturing sector (dummy) Yes = 1, No = 0 - 0.783 0.413

Dependent Variable:
Log hourly wage Natural log of hourly wage 2.341 0.384

Table 2: Regression of Personal and Firm Characteristics on
Foreign Workers' Hourly Wage: Hamamatsu

Independent Variable Coefficient Standard Error Significance

Female (dummy) -0.2993** 0.0464 0.0000
Age 0.0735 0.0825 0.3745
Age2 -0.0022 0.0024 0.3677
Age3/1000 0.0199 0.0229 0.3857
Years of schooling 0.0024 0.0072 0.7438
Speaks/understands Japanese well (dummy) 0.0230 0.0524 0.6607
Time (months) living in Japan -0.0037 0.0044 0.4039
(Time [months] living in the Japan)2 0.0001 0.0001 0.1493
(Time [months] living in the Japan)3/1000 -0.0004* 0.0003 0.0962
Entered Japan illegally (dummy) -0.1064 0.1077 0.3246
Found job through friend/relative (dummy) 0.2027** 0.0583 0.0006
Work through labor broker (dummy) 0.1714** 0.0634 0.0074
Years with current employer -0.0032 0.0136 0.8132
Nikkeijin 0.2472** 0.0682 0.0004
Construction sector (dummy) -0.2596 0.1675 0.1226
Manufacturing sector (dummy) -0.2716** 0.0663 0.0001
Constant 1.5772 0.8574 0.0673
Note: Coefficients are unstandardized ordinary least-square regression coefficients.
The industrial sector reference group is the service sector other.
Sample Size: 217
** significance p<0.05, *significance p<0.10
R square: 0.332
Adjusted R square: 0.279
Standard error of estimate: 0.3260

Source: Survey of 336 foreign workers employed in Hamamatsu, Japan, in April-June 1996, conducted by researchers affiliated with the Institute for Statistical Research (Tokyo), in collaboration with the Center for U.S.-Mexican Studies, University of California-San Diego.

Table 3: Definition and Characteristics of Variables Used in
San Diego Wage Determination Model

Variable Measurement Expected Sign Mean Std. Dev.

Female (dummy) Female = 1, Male = 0 - 0.330 0.471
Age Years + 34.590 10.174
Years of schooling Years + 8.591 4.042
Speaks English well (dummy) Yes = 1, No = 0 + 0.338 0.474
Time living in the US Months + 143.665 96.592
Found job through friend/relative (dummy) Yes = 1, No = 0 +,- 0.692 0.462
Entered U.S. illegally (dummy) Illegal = 1, Legal = 0 - 0.584 0.493
Years with current employer Years + 4.668 4.734
Annual turnover rate among production workers Percent - 14.835 16.866
Percentage of foreign workers in workforce Percent +,- 68.861 24.000
Need English at work (dummy) Yes = 1, No = 0 + 0.378 0.486
Natives apply to same kind of job (dummy) Yes = 1, No = 0 +,- 0.408 0.492
Agriculture sector (dummy) Yes = 1, No = 0 + 0.126 0.332
High tech manufacturing sector (dummy) Yes = 1, No = 0 + 0.094 0.292
Construction sector (dummy) Yes = 1, No = 0 + 0.091 0.288
Hotel & motel sector (dummy) Yes = 1, No = 0 + 0.083 0.276
Landscape and maintenance sector (dummy) Yes = 1, No = 0 + 0.107 0.310
Apparel and textile sector (dummy) Yes = 1, No = 0 + 0.121 0.326
Food processing sector (dummy) Yes = 1, No = 0 + 0.137 0.344
Low tech manufacturing sector (dummy) Yes = 1, No = 0 + 0.113 0.317
Miscellaneous services sector (dummy) Yes = 1, No = 0 + 0.035 0.184

Dependent Variable:
Log hourly wage Natural log of hourly wage 1.826 0.356

Table 4: Regression of Personal and Firm Characteristics on
Foreign Workers' Hourly Wage: San Diego County

Independent Variable Coefficient Standard Error Significance

Female (dummy) -0.0451 0.0322 0.1619
Age 0.0864** 0.0321 0.0074
Age2 -0.0023** 0.0008 0.0038
Age3/1000 0.0202** 0.0063 0.0016
Years of schooling 0.0079* 0.0043 0.0675
Speaks English well (dummy) 0.1087** 0.0360 0.0027
Time (months) living in the US 0.0026** 0.0010 0.0082
(Time [months] living in the US)2 -0.00001** 0.0000 0.0434
(Time [months] living in the US)3/1000 0.00001 0.0000 0.1192
Found job through friend/relative (dummy) -0.0835** 0.0302 0.0061
Entered U.S. illegally (dummy) 0.0042 0.0312 0.8934
Years with current employer 0.0147** 0.0034 0.0000
Annual turnover rate (%) among production workers -0.0029** 0.0009 0.0011
Percentage of foreign workers in workforce -0.0015** 0.0007 0.0380
Need English at work (dummy) 0.0830** 0.0356 0.0204
Natives apply to same kind of job (dummy) 0.0992** 0.0364 0.0067
Agriculture sector (dummy) 0.1991** 0.0653 0.0025
High tech manufacturing sector (dummy) 0.2167** 0.0713 0.0026
Construction sector (dummy) 0.4874** 0.0656 0.0000
Hotel & motel sector (dummy) 0.1745** 0.0691 0.0120
Landscape and maintenance sector (dummy) 0.2090** 0.0628 0.0010
Apparel and textile sector (dummy) 0.2158** 0.0614 0.0005
Food processing sector (dummy) 0.1810** 0.0625 0.0040
Low tech manufacturing sector (dummy) 0.2299** 0.0629 0.0003
Miscellaneous services sector (dummy) 0.2907** 0.0864 0.0009
Constant 0.4041 0.3993 0.3122
Note: Coefficients are unstandardized ordinary least-square regression coefficients.
The industrial sector reference group is the restaurant sector.
Number of cases: 373
** significance p<0.05, *significance p<0.10
R square: 0.500
Adjusted R square: 0.464
Standard error of estimate: 0.2604

Source: Survey of 501 immigrant workers employed in San Diego County, California, January-March 1996, conducted by the Center for U.S.-Mexican Studies, University of California-San Diego.