A Study of the U.S.-Mexico Border Earnings Penalty between 2000 and 2005

Marie T. Mora Professor of Economics Department of Economics and Finance College of Business Administration University of Texas—Pan American Edinburg, TX 78541 USA Fax: 011-956-384-5020; [email protected] Alberto Dávila Professor and V.F. “Doc” and Gertrude Neuhaus Chair for Entrepreneurship Department of Economics and Finance College of Business Administration University of Texas—Pan American Edinburg, TX 78541 USA Fax: 011-956-384-5020; [email protected] and André Varella Mollick Associate Professor of Economics Department of Economics and Finance College of Business Administration University of Texas—Pan American Edinburg, TX 78541 USA Fax: 011-956-384-5020; [email protected]

March 2007

A Study of the U.S.-Mexico Border Earnings Penalty between 2000 and 2005

Abstract: Several studies have identified a substantial “border penalty” on the earnings of individuals who reside in areas close to the U.S. - Mexico southern border. This paper uses IPUMS data based on the 2000 decennial census and the 2005 American Community Survey (ACS) 2005 samples to analyze whether this border penalty changed during the early 2000s. We first document that a border/interior earnings gap exists for the three groups of workers, despite higher average education levels within the same ethnic group along the border. Morerover, while the border earnings penalty is particularly pronounced among Mexican immigrants (-0.128), Mexican immigrants and non-Hispanic whites along the border experienced an earnings improvement between 2000 and 2005 compared to their counterparts in the rest of the U.S. (of approximately 4.4% for Mexican immigrants and 2.7% for non-Hispanic whites). Studying the manufacturing, construction and trade sectors in more detail, we find that the U.S. born MexicanAmericans are associated with a substantially higher border penalty if working in any of these sectors. The more stable border penalty premiums for Mexican immigrants (with the exception of construction (-0.213)), suggest a high degree of adaptation to the changing business cycles in the early 2000s. As a whole, our findings are consistent with at least two explanations: i) the more mobile Mexican immigrant population might be responding faster to changes in the U.S. business cycles; and ii) Mexican immigrants appear to be gaining ground versus MexicanAmericans in the job pool.

Deleted: Compared to U.S.-born Mexican Americans and to Non-Hispanic Whites, this border penalty is particularly pronounced among Mexican immigrants. Deleted: on metropolitan statistical areas (MSAs) bordering Mexico based on the 2000 IPUMS and the Deleted: Focusing on a very short time span, we are interested in checking whether the business cycles of the early 2000s have had any systematic impact on any group of individuals. Deleted: in Deleted: W Deleted: “ Deleted: ” Deleted: , however, other than those working in Deleted: very good Deleted: ’ Deleted: stories Deleted: is Deleted: are perhaps

2

A Study of the U.S.-Mexico Border Earnings Penalty between 2000 and 2005 1. Introduction Underlying economic and social characteristics along the U.S. - Mexican border, that in many cases differ greatly from those of the U.S. interior, have made this region an important area of study. For example, the effects of the large influx of mostly Mexican immigrants into U.S. states and the degree of dependence between Mexican and U.S. economic growth are usually framed first by considering potential impacts along the border region. Most recently, the U.S. economic slowdown in 2001 is said to have contributed to a slowdown of Mexican exports to the U.S. and the reduction of maquiladora activity taking place in several of the Mexican northern Deleted:

states, as documented by Coronado et al. (2004) for the Tijuana-San Diego area and by Mollick, and Wvalle-Vázquez (2006) for other northern Mexican states. Among the most documented impacts of the aforementioned effects is the existence of the “border penalty”, the stylized fact that workers along the U.S.-Mexico border earn less on average than those in the U.S. interior. At least two theories have been put forth to explain it. First, Mexican immigrant labor’s supply could be highly elastic, depressing wages along the U.S.-Mexico border, as first suggested empirically by Smith and Newman (1977). In pursuing this explanation, more recent studies by Mora (2006) and Mora and Dávila (2006) have pointed to the importance of distinguishing between U.S.-natives and Mexican immigrants, and have provided evidence of significant differences in the border-earnings disparity according to immigration status. A second theory argues that earnings are lower in this region because of relatively low average levels of human capital. Evidence found in Mora (2006), Mora and

1

Formatted: Font: 12 pt Deleted:

Dávila (2006), Flota and Mora (2001), Fullerton (2001), Dávila and Mora (2000), Peach (1997), Dávila and Mattila (1985), and Dávila (1982) provides support for this theory. This second theory, however, does not elaborate on why workers along the U.S.-Mexico border have lower levels of human capital than in the U.S.-interior. One explanation might relate to the dependence of the border economy on the Mexican economy and, as such, to macroeconomic shocks emanating from both Mexico and the U.S. A well known set of empirical findings documents that workers with different characteristics do not similarly respond to economic shocks. For example, studying 30 years over the 50 Spanish provinces, Mauro and Spilimbergo (1999) find that highly skilled workers migrate very promptly after a decline in regional labor demand, while low-skilled workers drop out of the labor force or remain unemployed. With regards to the border case, do adverse economic shocks disproportionately drive skilled labor away from the border? Also, this possibility conceptually suggests an uneven effect on the earnings of residents of this region. For example, a decrease in the skill level of workers or capital investments along the border might negatively impact the earnings position of labor serving as a complement to this production input. This conceptualization of the border earnings penalty is short-run in nature, however, and must be tempered by long-run dynamics. That is, according to neoclassical theory, regional earnings differentials should disappear over the long run for at least three reasons, including: labor migration from low wage areas to high wage areas (labor supply effect); gravitation of entrepreneurs and capital to those areas with relatively lower labor costs (labor demand effect); and a consumer-demand effect, in which those regions with lower earnings produce goods and services at lower prices, and would therefore attract consumer demand for its products.

2

With these conceptual points in mind, newly released data might shed more light on the underpinnings of the border penalty. In particular, in this paper we empirically estimate this border penalty over a short time span, employing data from the 2000 U.S. decennial census and the 2005 American Community Survey (ACS), which are available in the Integrated Public Use Microdata Series (IPUMS). These samples encompass a time period that saw important socioeconomic changes along the border and elsewhere. For one, immigration to the U.S. along its southern border intensified during this time, opening up the possibility to further test the first theory mentioned above. For another, there was a slow-down in maquiladora activity in the northern Mexico at the same time that the construction industry flourished along the U.SMexican border, allowing us to test for the foregoing possibility that a “skill drain” might have accentuated this penalty. The border wage differentials are estimated by first comparing the mean earnings differentials between border and interior workers across three groups: Mexican Americans, Mexican immigrants and non-Hispanic whites. We then estimate these differentials by controlling for a host of socioeconomic characteristics of workers in a fully interactive earningsfunction setting. Finally, we estimate these earnings gaps across the wage distribution of workers by using earnings-quantile analyses. Our major results are as follows. We first document that a border/interior earnings gap exists for the three groups of workers, despite higher average education levels along the border when comparing workers of the same ethnicity. While the border earnings penalty is particularly pronounced among Mexican immigrants (nearly 13 percent), Mexican immigrants and nonHispanic whites along the border experienced an earnings improvement between 2000 and 2005

3

compared to their counterparts in the rest of the U.S. (of over four percentage points for Mexican immigrants, and nearly three for non-Hispanic whites). Studying key industrial sectors in more detail, we find that U.S. born Mexican Americans working in manufacturing, construction and trade any of these sectors accrue a substantially higher border penalty than the average case. The more stable border penalty premiums for Mexican immigrants, other than those working in construction, suggest a high degree of adaptation to the changing business cycles in the early 2000s. This observation is consistent with the more mobile Mexican immigrant population responding faster to changes in U.S. business cycles, and with Mexican immigrants gaining ground versus Mexican-Americans in the job pool.

2. The Data and Preliminary Findings Our study of the U.S. border earnings penalty employs data from the 2000 decennial

census and the 2005 ACS, made available by Ruggles et al (2007) in the IPUMS. Our sample includes U.S.- and foreign-born Mexican Americans between the ages of 25 and 64 who worked for wages and salaries for at least 20 hours per week for 32 weeks in the previous year. For the sake of comparison, we also consider the conventional base-group of U.S.-born non-Hispanic whites. Individuals reporting military occupations are excluded from our sample. While the information on earnings pertains to the year prior to the surveys (i.e., for 1999 and 2004), for ease of discussion, we refer to the years in which the surveys were conducted. Following Mora and Dávila (2007), we define the U.S. border region using the public use microdata areas (PUMAs) for county clusters available in the IPUMS; PUMAs containing counties adjacent to Mexico represent the U.S. border region. Specifically, our border definition includes Cameron, Hidalgo, Starr, Jim Hogg, Brooks, Willacy, Kenedy, Kleberg, Zapata, Webb, La Salle, Dimmit, Zavala, Uvalde, Maverick, Kinney, Real, Edwards, Val Verde, Terrell, Pecos,

4

Brewster, Presidio, Jeff Davis, Reeves, Upton, Crane, Ward, Winkler, Loving, Andrews, Gaines, Culberson, Hudspeth, and El Paso Counties in Texas; Doña Ana, Otero, Chaves, Eddy, Lea, Luna, Hidalgo, Grant, Sierra, Socorro, Torrance, and Catron Counties in New Mexico; Cochise, Santa Cruz, Graham, Greenlee, Pima, and Yuma Counties in Arizona; and Imperial and San Diego Counties in California. We opt for this approach instead of using metropolitan statistical areas (MSAs—cities with populations of at least 100,000 residents) to identify the border in this study because some U.S. border communities too small to be identified as MSAs have relatively large Mexican “sister” cities, and thus depend heavily on cross-border economic activities.1 In the 2000 IPUMS, PUMAs are only identified in the 5% version. As such, our 2000 sample is drawn from the 5% IPUMS sample for residents of the four border states (Arizona, California, New Mexico, and Texas), and the 1% IPUMS sample for workers in other states. The 2005 ACS, despite being a 1% sample, includes the PUMAs. In all of our analyses, we employ the appropriate sampling weights provided by the IPUMS to maintain the national representation of the data. Table 1 provides the average characteristics of Mexican immigrants, U.S.-born Mexican Americans, and U.S.-born non-Hispanic whites residing along the U.S.-Mexico border and those in the interior. Earnings are defined as the annual wage and salary income divided by usual weekly work-hours times weeks worked. Note that for workers of the same ethnicity and birthplace, those along the border have higher average education levels; yet, a border/interior wage gap exists for Mexican immigrants (a 10.1 percent gap in 2000) and U.S.-born Mexican Americans (14.6 percent). This finding indicates that low average wages along the U.S.-Mexico border are not unique to immigrants and do not solely reflect differences in human capital. 1 For example, as noted by Mora and Dávila (2007), Eagle Pass, Texas and Del Rio, Texas have less than 100,000 residents, but each of their Mexican “sister” cities (Piedras Negras and Ciudad Acuña) have more than 100,000 residents. International bridges connect the city pairs, and are open 24 hours per day.

5

[Insert Table 1 about here] Of specific interest to this study are changes in these earnings differentials between 2000 and 2005. Note that for Mexican immigrants, the border-earnings penalty narrowed from 10.1 percent in 2000 to 4.7 percent in 2005; this difference is statistically significant at the onepercent level. However, U.S.-born Mexican Americans did not experience a similar (or Deleted: <
significant) decline in the border-earnings gap. This change was not statistically significant for US-born Mexican-Americans. This result suggests that, compared to their counterparts in the U.S.-interior, Mexican immigrants in areas close to Mexico gained ground with respect to

Formatted: Highlight Formatted: Highlight Formatted: Highlight Formatted: Highlight Deleted:

earnings relative to their U.S.-born peers. Of interest, U.S.-born non-Hispanic whites living

Deleted: ?

close to Mexico earned more on average than their counterparts in the rest of the U.S. in both

Deleted: I didn’t think it was—we need to check>>

years, particularly in 2005.

Formatted: Highlight

Formatted: Highlight

Formatted: Highlight

Trejo (1997) compares 1979 to 1989 real wage data taken from CPS surveys. He finds a very contrasting pattern between Mexican-origin workers and Mexican Americans: in 1979, mean log wages of Mexicans were 0.261 below those of non-hispanic whites and 0.023 below those of blacks. By 1989, the Mexican wage disadvantage had widened to 0.35 log points relative to whites and 0.052 relative to blacks. On the other hand, the outlook was much better for U.S. born Mexican-Americans: in 1979, mean log wages were 0.151 below those of whites and 0.07 above those of blacks; in 1989, the corresponding wage differences were -0.239 and 0.045. In other words, in a single decade in which its first year coincided with the onset of a recession the wage gap increased for Mexican immigrants and Mexican Americans alike. In Trejo’s work the two surveys analyzed were conducted at very similar phases of the business cycles: at the start of recessions.

6

Deleted: data

Some of the earnings improvement for foreign-born versus U.S.-born Mexican Americans in cities near Mexico could stem from the larger increase in the education levels of immigrants between 2000 and 2005, a fact already noted by Dávila and Mora (2006) for the decade 1990 and 2000. The average schooling of Mexican-born workers residing close to their home country rose by 0.8 years (from 9.6 to 10.4 years), which is slightly more than the 0.7-year increase for those in the U.S.-interior. The increase in the average education of U.S.-born Mexican Americans and non-Hispanic whites was smaller and similar between the border and interior regions (0.1 years for U.S.-born Mexican Americans, and 0.2 or 0.3 years for nonHispanic whites). Additional observations in Table 1 are consistent with extant work in Mora and Dávila (2006a) and Dávila and Mora (2000) in that, compared to those in the U.S.-interior, Mexican immigrants along the border tend to be more “established” in the U.S. and possess greater levels of human capital. On average, Mexican workers in U.S.-border communities have more education, more years of potential work experience (measured by the convention of age minus education minus 5 years), and are less likely to be limited-English-proficient (LEP) [a binary variable equal to one for individuals who did not speak English “well”, and equal to zero otherwise.]

3. Empirical Methodology We consider next the extent to which these observable differences explain changes in the immigrant/native earnings differentials between 2000 and 2005. We first estimate a standard Mincerian-type earnings function:

ln (W) = (Border) α1 + (Border x 2005IPUMS) α2 + X β1 + (X x 2005IPUMS) β2 + regional dummies β3 + ε (1),

7

where: ln (W) represents the natural logarithm of hourly earnings. The vector Border includes a binary variable equal to one for workers in U.S.-border cities (= 0 otherwise), while (Border x 2005IPUMS) denotes the binary Border variable interacted with a binary variable equal to one for individuals in the 2005 IPUMS (= 0 for 2000). As such, the coefficient vector α1 reflects the earnings “penalty” associated with residing along the U.S.-Mexico border, while α2 measures changes in this penalty between 2000 and 2005.2 X is a vector of additional variables that affect earnings, including education, experience, experience-squared, limited-English-proficiency (for immigrants), immigrants’ time in the U.S., gender, and a constant term. The interactive variable X x 2005IPUMS represents a vector of the variables in X interacted with 2005IPUMS. The equations were estimated with 5 regional dummy variables as additional controls (out of 6 regional dummies: New England, Mid-Atlantic, North Central, South Central, and Mountain), together with their respective interactive terms with 2005 IPUMS. The omitted category is the Pacific region. These 10 additional coefficients are omitted for space constraints but capture regional effects in the estimations. Finally, ε is the normally distributed error term. After estimating (1) by OLS, we then consider whether differences existed between Deleted: the

workers in particular industries, such as manufacturing. Finally, we explore conditional Deleted: s

quantile regressions as surveyed by Koenker and Hallock (2001) to provide additional insights Deleted: testing

into whether workers in the low end of the earnings distribution experienced a similar earnings improvement as those at the high end.

4. Results 2

For interpretative ease, we will discuss the estimated αi’s as earnings penalties, although a more precise interpretation exists (Kennedy 1981). 8

Table 2 contains the results from estimating Equation (1) separately for Mexican immigrants, U.S.-born Mexican Americans, and U.S.-born non-Hispanic whites. In all three cases, consistent with the border literature, residing in cities along the U.S.-Mexico border relates to significantly lower earnings controlling for other observable characteristics. This border earnings penalty is particularly pronounced among Mexican immigrants (nearly 13 percent), although perhaps less severe than values previously reported. For example, Smith and Newman (1977) employed data based on the 1970 Census and reported a roughly 20% annual nominal wage differential and 8% annual real income differentials. They interpreted the latter as the “implicit premium that individuals along the border are willing to pay for nonpecuniary advantages such as remaining close to their cultural heritage.” Smith and Newman (1977, p. 63). Using data from 1990, Dávila and Mora (2006) suggest a near 19-percent border penalty among Mexican immigrants. [Insert Table 2 around here] However, when focusing on the first and third columns, Mexican immigrants and nonHispanic whites along the border experienced a significant earnings improvement between 2000 and 2005 compared to their counterparts in the rest of the U.S. (of approximately 4.4 percentage points for Mexican immigrants, and 2.7 points for non-Hispanic whites). On the other hand, the relative earnings of Mexican Americans along the border did not significantly improve over the five-year period. These earnings gains suggest a positive effect to Mexican immigrants residing close to their home country in the 2000s beyond changes in their observable skill levels. The positive effect is also observed to a certain extent for non-Hispanic whites and could be attributed to the loosening of monetary policy from 2001 to 2004.3 Anecdotal evidence suggests

Deleted: <
Formatted: Highlight 3

As a robustness test, we also performed Oaxaca (1973)-type wage decompositions by first estimating Equation (1) for each of the three groups confined to the non-border sample. We then used these estimated coefficients to predict 9

Deleted: NHWs

the employment of non-Hispanic whites as managers of maquiladoras increased during this

Formatted: Highlight Formatted: Highlight

time.

Deleted: —anecdotal evidence suggests it did>>

Other coefficients have the expected signs on earnings. Education and experience have

Formatted: Highlight

positive effects, while experience squared has small negative effects. The impact of the education coefficient on earnings is larger for non-Hispanic whites (0.110) than for Mexican Americans (0.088), followed by Mexican immigrants (0.026), which suggests lower returns to education to the latter type of workers. This could reflect the fact that some of the Mexican immigrants’ skills were acquired abroad, as developed by McManus et al. (1983). The LEP variable has a negative impact on Mexican immigrants and an increase in years in the U.S. contributes, ceteris paribus, to higher earnings for this same group of workers as expected. Note also the systematic negative effects associated with female workers (varying from -0.210 for Mexican Americans to -0.304 for non-Hispanic Whites), which is a well established fact in the literature. In short, Table 2 indicates that, despite the existence of a border-earnings penalty, the average earnings of Mexican immigrants and U.S.-born non-Hispanic whites improved along the border vis-à-vis the U.S.-interior, ceteris paribus, during the five-year period. U.S.-native Mexican Americans working in the U.S. border region did not experience a reduction in the

Deleted: XX

Formatted: Highlight Deleted: YY

border penalty. One possible explanation for our findings is that Mexican immigrants are more

Formatted: Highlight Deleted: ZZ

mobile than Mexican Americans, in which the former are more predisposed to move to areas that offer more attractive earnings. Another possibility is the attractiveness for the entrepreneurs and the earnings of workers along the border had they faced the same structure of wages as their non-border counterparts. The difference between the actual average earnings and these predicted values in each ethnic group provides another estimate of the border-wage penalty. In 2000, these border penalties were 13.4 percent for Mexican immigrants, 10.7 for U.S.-born Mexican Americans, and 4.7 for non-Hispanic whites. However, by 2005, the magnitude of these penalties significantly declined for Mexican immigrants (to 8.9 percent), and for nonHispanic whites (to 1.9 percent). For U.S.-born Mexican Americans workers, however, the 2005 border earnings penalty of 10.1 percent was not statistically different at conventional levels from their 2000 penalty. In sum, these wage-decomposition findings support the OLS results reported in Table 2. 10

Formatted: Highlight Deleted: XX

Formatted: Highlight Formatted: Highlight Formatted: Highlight Deleted: YY

Formatted: Highlight Formatted: Highlight Deleted: XX

Formatted: Highlight

capital when the Mexican immigrant population has the familiarity and skills to more efficiently engage in trade with Mexican consumers and businesses, as discussed by Mora and Dávila (2006). Moreover, based on data from the 1970 Census of four Texas MSAs (Brownsville, Laredo, Corpus Christi and Houston), Smith and Newman (1977) found numbers supportive to the notion that legal and illegal migration impact resident Mexican-Americans the most. Deleted: ¶

The Border Earnings Penalties in Specific Industries. Table 3 investigates further whether workers in specific industries had disproportionate responses to the geographic location

Formatted: Font: Italic Deleted: activity Deleted: ve

or to worker’s characteristics. We are particularly interested in the manufacturing, construction Deleted: workers

and trade sectors. This choice of industry follows from the respective declines in manufacturing Deleted: very much

in overall U.S. output, the sensitive housing sector to business cycles, and to the compensating

Deleted: the

dynamics of the trade sector in the border region, as documented by Adkisson and Zimmerman (2004).4 [Insert Table 3 around here] Deleted: the

Compared to the baseline cases reported in Table 2, U.S. born Mexican-Americans Deleted: are associated with

accrued a substantially higher border penalty when working in any of these sectors compared to

Deleted: if

the -0.106 coefficient reported in the general case. In column 2 of Table 3, for example, the border-penalty coefficient for U.S. born Mexican-Americans is -0.161 for manufacturing, -0.162 for construction activities, and -0.129 for trade workers. For Mexican immigrants, the borderDeleted: is about the same for

penalty coefficients in manufacturing and trade (-0.122 and -0.130), are similar to the -0.128 coefficient value reported in Table 2 for the entire sample, but are smaller in magnitude than the

4

Adkisson and Zimmerman (2004) examine changes in border-region general retail, clothing and accessory, grocery, and furniture and home furnishings sales that occurred during the 1992-1997 NAFTA implementation era. They find, ceteris paribus, that retail sales in border MSAs is larger in proportion to local income than they are in non-border MSAs. After controlling for other factors, their results suggest a negative influence on retail sales on the U.S. side of the border for the 1992 and 1997 period, although the changes cannot be unquestionably attributed to NAFTA. 11

Deleted: sector workers Deleted: or Deleted: which are very close to the

Deleted: of -0.128, but much higher for

penalty earned by construction workers (-0.213). That is, foreign-born construction workers Deleted: This suggests a much

accrued a higher penalty along the border than elsewhere. For non-Hispanic Whites,

Deleted: for workers in construction and located in the border

manufacturing workers are associated with a smaller border penalty (-0.025). The lower border

Deleted:

penalty for non-Hispanic whites in manufacturing can be contrasted to those in construction (0.045) or trade sectors (-0.041) which more closely resemble the average in Table 2 of -0.048 for

Deleted: <
Formatted: Highlight Deleted: much

such workers.

Deleted: >>>>>>>¶

Table 3 further shows that the earnings of foreign-born and U.S.-born Mexican American construction workers along the border significantly improved by around ten percentage points

Deleted: Two more findings are evident from the sector-based approach. First, Deleted: for

the border. The border region is a rapidly growing area with higher rates of population

Deleted: of either Mexican immigrant or Mexican-Americans, there is a positive effect given by the interactive dummy for the 5-year period and the border MSA. The positive effect on earnings due to this factor varies from 0.104 for Mexican immigrants to 0.098 for MexicanAmericans, values much larger than those reported in Table 2

growth than other regions, which might explain why construction workers seemed to

Deleted: positive effects of the loosening of monetary policy

(see the coefficient on the Border x 2005 IPUMS interaction). This suggests that the housingmarket boom after 2001 seemed to have particularly benefited Mexican American workers along

Deleted: probably

particularly gain along the border.5

Deleted: these two categories of workers

Quantile Regression Analysis. Our results thus far indicate that the border wage penalty declined on average for Mexican immigrants and non-Hispanic whites between 2000 and 2005; however, this decline did not seem to occur “across the board” with respect to industry. It is also

Deleted: <
Formatted: Highlight Formatted: Highlight Deleted: —this

Formatted: Highlight

of interest to explore how the border-earnings penalty evolved for workers in the low end of the

Deleted: r?>>

earnings distribution vis-à-vis those at the high end.. One particular technique, conditional

Deleted: Second, Table 3 also shows the returns to education in these regressions for the three sectors of activity. In all cases, in agreement to what is expected, construction workers have less favorable impacts to an additional ... [1]

quantile regressions as reviewed by Koenker and Hallock (2001), provides insight into this issue. We therefore estimate Equation (1) for nine earnings deciles in each year for each of the

Formatted: Highlight

Deleted: Also, I’m not sure if we ... [2]

Formatted: Highlight 5

López (2006) discusses the Rio Grande Valley (southern-eastern part of Texas) recent growth in detail. The McAllen MSA posted the strongest gains of all the Texas-Mexico border metros from 1997 to 2003, with employment growing an average of 4.6% and 3.1% in Brownsville’s MSA, against the 1.6% job growth of Texas or of the whole U.S. (1.2%). Cañas et al. (2005) emphasize the extent to which local Texas border MSA growth in the 1990-2000 period depends on the Mexican maquiladora growth. They mention, in particular, that “maquiladora employees draw their salary in Mexico but do a significant share of their shopping in the United States, stimulating employment in local retail and service sectors.” Cañas et al. (2005, p. 29). 12

Formatted: Font: Bold, Highlight Deleted: It is interesting to know ... [3] Deleted: the Deleted: additional Deleted: s. We test whether workers in ... [4] Deleted: the

Deleted: ,

three ethnic/native samples; Figures 1, 2 and 3 display the estimates of the coefficients on the Deleted: for the three groups of interest

Border variable. For reference purposes, we report the OLS-based coefficients from Table 2 as Deleted: ¶

the straight line. Figure 1 shows for Mexican immigrants, that the improvement has been particularly visible for top-end earnings workers. For both 2000 and 2005 quantile estimations, the coefficients associated with quantiles 0.8 or 0.9 are especially lower (in absolute value) than the OLS one of -0.126: -0.110 for quantile 0.8 and -0.086 for quantile 0.9 in 2000, in contrast to 0.076 for quantile 0.8 and -0.045 for quantile 0.9 in 2005. Figure 2 contains the results for Mexican Americans, displaying a flatter profile: there is an improvement over the 5-year period but of much smaller magnitude than the gains obtained for Mexican immigrants. For U.S-born Mexican Americans, the mid-quantiles coefficients are especially close to the OLS ones, while more dispersion is noticed at the lower and higher end of earnings distributions. For nonHispanic Whites in Figure 3, the positive shift over the 5-year period is very pronounced, although less so for those workers in the top end of the distribution (0.8 and 0.9 deciles). The evidence from quantile regressions thus reinforces our findings that the only category that did not have earnings improvement was the Mexican-American workers. While the improvement in the 5-year period was more evident for top earners for Mexican immigrants, it was less clear so for non-Hispanic Whites. [Insert Figures 1, 2, 3 here] At least in another dimension the whole distribution of earnings of Mexican-Americans is is sharp contrast to those of Mexican-immigrants. As documented by Buchinsky (1994) and others, education has a greater effect upon the wages of individuals at the top of the wage distribution. We observe this general finding by employing quantile regressions and investigating

13

the education coefficient in the earnings regression. For Mexican-immigrants, we indeed find in Figure 4 a very symmetric function (with center of gravity in the 5th quantile) with high-wage earners experiencing a much higher effect of returns to education on earnings than for low wage earners. A similar, although less impressive pattern, is found for non-Hispanic Whites, while the returns on education for Mexican-Americans clearly violate this general rule. The full results from quantile regressions on the returns to education are available upon request from the authors. [Insert Figure 4 here]

5. Concluding Remarks Adopting a Mincerian type earnings framework, this research complements the empirical works with regional labor demand effects in Mauro and Spilimbergo (1999), the varying response of monetary policy to groups of workers in Zavodny and Zha (2000). It also sheds light on differentiated effects of business cycles on workers under the calibration model in Mukoyama and Sahin (2006). We first document that a border/interior earnings gap exists for the three groups of workers, despite higher average education levels in U.S. MSAs near Mexico. This “border earnings penalty” is particularly pronounced among Mexican immigrants (-0.128) after accounting for worker’s characteristics. However, Mexican immigrants and non-Hispanic whites along the border experienced an earnings improvement between 2000 and 2005 compared to their counterparts in the rest of the U.S. (approximately 4.4% for Mexican immigrants and 2.7% for non-Hispanic whites). If there is any pecuniary (negative earnings) effect to staying close to the border, this research suggests this premium was partly offset during the 5-year period for Mexican immigrants and, less so, for non-Hispanic whites. Given the fact that the time period adopted encompasses both a monetary tightening (from 1999 to 2000) and then loosening following the 2001 recession, we study in detail the

14

manufacturing, construction and trade sectors. Mexican immigrants compete primarily with U.S. born Mexican-American workers in the environment following the negative shocks of 2001 due to recession and the terrorist attacks to the U.S. We find that the U.S. born Mexican-Americans are associated with a substantially higher border penalty if working in any of these sectors compared to the -0.106 coefficient reported in the general case. The fact that the border-penalty coefficients are about the same for manufacturing and trade sector for Mexican immigrants workers (-0.122 or -0.130) and very close to the total Mexican immigrants (-0.128), suggests a very good degree of adaptation to the changing business cycles, other than construction workers (-0.213). One consistent story that comes from these results is that the more mobile Mexican immigrant population responds faster to changes in U.S. business cycles. Another is that Mexican immigrants are gaining ground versus Mexican-Americans in the job pool. We confirm the general finding in Buchinsky (1994) of top wage earners experiencing higher returns to education by employing “quantile regressions” for Mexican-immigrants and less so for non-Hispanic Whites. The returns on education for Mexican-Americans, however, clearly violate this general rule. Exploration of these issues is left for further research, including alternative methods for decomposing changes in wage distributions proposed by Machado and Mata (2005) for quantile regressions.

15

References Adkisson, Richard, and Linda Zimmerman. 2004. “Retail Trade on the U.S-Mexico Border during the NAFTA Implementation Era.” Growth and Change 35 (1): 77-89. Buchinsky, Moshe. 1994. “Changes in the U.S. Wage Structure 1963-1987: Application of Quantile Regression.” Econometrica 62 (2): 405-458. Cañas, Jesús, Roberto Coronado, and Robert Gilmer. 2005. “Texas Border: Employment and Maquiladora Growth.” FRB of Dallas, October: 27-32.

Formatted: Indent: Left: 0", First line: 0"

Coronado, Roberto, Thomas Fullerton, and D. Clark. 2004. “Short-run Maquiladora Employment Dynamics in Tijuana.” Annals of Regional Science 38: 751-763.

Deleted: .

Dávila, Alberto, and Marie T. Mora. 2006. “Changes in the Relative Earnings Gap between Natives and Immigrants along the U.S.-Mexico Border.” Paper presented at the Lineae Terrarum International Borders Conference, Ciudad Juárez, March 2006. Dávila, Alberto, and Marie T. Mora. 2000. “English Skills, Earnings, and the Occupational Sorting of Mexican Americans Working along the US-Mexico Border.” International Migration Review 34:133-57.

Formatted: Indent: Left: 0", First line: 0.5"

Deleted: T. Deleted: , D. Deleted: Deleted: ¶

Formatted: Spanish (Spain-Modern Sort) Formatted: Spanish (Spain-Modern Sort)

Dávila, Alberto, and J. Peter Mattila, 1985. “Do Workers Earn Less along the U.S.-Mexico Border?” Social Science Quarterly 66:310-18. Dávila, Alberto. 1982. “Sources of Depressed Earnings Along the Texas-Mexico Border.” Federal Reserve Bank of Dallas Economic Review, November 1982. Flota, Chrystell, and Marie T. Mora. 2001. “The Earnings of Self-Employed Mexican Americans along the US-Mexico Border.” The Annals of Regional Science 35:483-99. Fullerton, Thomas M. 2001. “Educational Attainment and Border Income Performance.” Federal Reserve Bank of Dallas Economic and Financial Review, Third Quarter, pp. 2-10. Koenker, Roger, and Kevin Hallock. 2001. “Quantile Regression.” Journal of Economic Perspectives 15(4):143-156. López, José Joaquin. 2006. “Dynamic Growth in the Rio Grande Valley.” FRB of Dallas Southwest Economy, March-April: 11-14. Machado, José, and José Mata. 2005. “Counterfactual Decomposition of Changes in Wage Distributions Using Quantile Regressions.” Journal of Applied Econometrics 20: 445465.

Formatted: Spanish (Spain-Modern Sort)

Mauro, Paolo, and Antonio Spilimbergo. 1999. “How do the Skilled and the Unskilled Respond to Regional Shocks? The Case of Spain.” IMF Staff Papers 46 (1): 1-17.

Formatted: Spanish (Spain-Modern Sort)

16

McManus, Walter S., Gould, William, and Finis Welch. 1983. “Earnings of Hispanic Men: The Role of English Language Proficiency.” Journal of Labor Economics 1(2):101-130. Mollick, André, and Karina Wvalle-Vázquez. 2006. “Chinese competition and its effects on Mexican Maquiladoras.” Journal of Comparative Economics 34: 130-145. Mora, Marie T. 2006. “Self-Employed Mexican Immigrants Residing along the U.S.-Mexico Border: The Earnings Effect of Working in the U.S. versus Mexico.” International Migration Review 40(4): 885-98. Mora, Marie T. 2005. “Changes in Occupational Earnings along the US-Mexico Border between 1900 and 1920.” Journal of Economic Issues 39(4): 1043-59. Mora, Marie T., and Alberto Dávila. 2007. “Cross-Border Earnings of U.S.-Native Non-Hispanic Whites along the U.S.-Mexico Border.” The University of Texas – Pan American, Department of Economics and Finance, unpublished manuscript, February. Mora, Marie T., and Alberto Dávila. 2006. “Mexican Immigrant Self-Employment along the U.S. Mexico Border: An Analysis of 2000 Census Data.” Social Science Quarterly 87(1): 91-109. Mukoyama, Toshihiko, and Aysegul Sahin. 2006. “Costs of Business Cycles for Unskilled Workers.” Journal of Monetary Economics 53: 2179-2193. Oaxaca, Ronald. 1973. “Male-Female Differentials in Urban Labor Markets.” International Economic Review 14: 693-709. Peach, James T., and Richard V. Adkisson. 2000. “NAFTA and Economic Activity along the U.S.- Mexico Border.” Journal of Economic Issues 34: 481-89. Peach, James T., 1997. “Income Distribution along the United States Border with Mexico.” Journal of Borderlands Studies 12:1-15. Ruggles, Steven, and Sobek, Matthew, et al. 2003. Integrated Public Use Microdata Series, Version 3.0, University of Minnesota, Minneapolis: Historical Census Projects; http://www.ipums.org . Smith, Barton, and Robert Newman. 1977. “Depressed Wages along the US-Mexico Border.” Economic Inquiry 15: 51-66. Trejo, Stephen J. 1997. “Why do Mexican Americans Earn Low Wages?” Journal of Political Economy 105 (6): 1235-1268.

17

Welch, Finis. 2000. “Growth in Women’s Relative Wages and in Inequality among Men: One Phenomenon or Two?” American Economic Review: AEA Papers and Proceedings 90(2): 444-49. Zavodny, Madeline, and Tao Zha. 2000. “Monetary Policy and Racial Unemployment Rates.” Federal Reserve Bank of Atlanta Economic Review, Fourth Quarter 2000: 1-16.

18

Figure 1: Estimated “Border Penalty” by OLS and by Quantile Regressions (2000 and 2005 coefficients) for Mexican Immigrants.

Earnings Quantiles 0 1

2

3

4

5

6

7

-0.02

-0.04

-0.06

-0.08

-0.1

-0.12

-0.14

-0.16 2000 coef.

2005 coef.

19

ols

8

9

Figure 2: Estimated “Border Penalty” by OLS and by Quantile Regressions (2000 and 2005 coefficients) for Mexican Americans.

Earnings Quantiles 0 1

2

3

4

5

6

7

-0.02

-0.04

-0.06

-0.08

-0.1

-0.12

-0.14 2000 coef.

2005 coef.

20

ols

8

9

Figure 3: Estimated “Border Penalty” by OLS and by Quantile Regressions (2000 and 2005 coefficients) for Non-Hispanic Whites.

Earnings Quantiles 0 1

2

3

4

5

6

7

-0.01

-0.02

-0.03

-0.04

-0.05

-0.06

-0.07

-0.08 2000 coef.

2005 coef.

21

ols

8

9

Figure 4: Estimated Education Coefficients by OLS and by Quantile Regressions (2000 and 2005 coefficients) for Mexican Immigrants.

Earnings Quantiles 0.045

0.04

0.035

0.03

0.025

0.02

0.015

0.01

0.005

0 1

2

3

4

5

2000 coef.

22

6 2005 coef.

7 ols

8

9

Table 1: Average Characteristics of Mexican Immigrants, Mexican Americans, and Non-Hispanic Whites along the U.S.-Mexico Border and in the U.S.-Interior in 2000 and 2005. U.S.-Born Mexican Americans

Mexican Immigrants Characteristic

2000

2005

U.S.-Born Non-Hispanic Whites

2000

2005

2000

2005

Residents of U.S. MSAs along Mexican border: Natural log hourly Earnings

2.143 (0.617)

2.311 (0.608)

2.397 (0.600)

2.553 (0.638)

2.802 (0.675)

3.041 (0.734)

Education

9.619 (4.266)

10.407 (3.757)

12.659 (2.700)

12.792 (2.618)

14.281 (2.338)

14.504 (2.342)

Experience

25.886 (11.217)

26.063 (10.935)

21.310 (10.197)

22.192 (10.662)

23.531 (10.082)

24.735 (10.529)

Years in U.S.

19.856 (10.814)

20.640 (11.983)

LEP

0.412 (0.492)

0.490 (0.500)

0.029 (0.169)

0.039 (0.193)

Female

0.372 (0.483)

0.385 (0.487)

0.440 (0.496)

0.450 (0.498)

0.421 (0.494)

0.420 (0.494)

N (unweighted)

13,494

2,980

13,593

3,215

33,013

7,104

N (weighted)

278,700

354,971

285,954

366,124

677,352

656,154

Residents of U.S. Interior MSAs: Natural log hourly Earnings

2.244 (0.565)

2.358 (0.550)

2.543 (0.592)

2.695 (0.604)

2.734 (0.659)

2.917 (0.679)

Education

8.849 (4.215)

9.552 (3.752)

12.547 (2.576)

12.681 (2.468)

13.793 (2.391)

13.970 (2.379)

Experience

23.456 (10.543)

23.453 (10.448)

21.419 (10.214)

21.450 (10.573)

23.762 (10.272)

24.788 (10.544)

Years in U.S.

16.220 (9.881)

16.422 (10.558)

LEP

0.473 (0.499)

0.524 (0.499)

0.023 (0.151)

0.031 (0.174)

Female

0.280 (0.449)

0.270 (0.444)

0.452 (0.498)

0.440 (0.496)

0.446 (0.497)

0.451 (0.498)

N (unweighted)

96,151

29,640

69,749

21,037

1.0e+06

682,017

2,736,995

4,228,383

1,875,676

2,554,085

64,279,101

63,520,851

N (weighted)

Notes: The parentheses contain the standard deviations for the continuous variables. These statistics use the IPUMS-provided sampling weights to maintain the national representation of the sample. The samples include workers employed for at least 20 hours per week for a minimum of 32 weeks in the previous year, and who resided in a metropolitan statistical area.

23

Table 2: Earnings Regression Results for Mexican Immigrants, U.S.-Born Mexican Americans, and U.S.Born Non-Hispanic Whites (Dependent Variable = Natural Logarithm of Hourly Earnings). Mexican Immigrants

U.S.-Born Mexican Americans

U.S.-Born NonHispanic Whites

U.S.-Mexico Border MSA

-0.128*** (0.006)

-0.106*** (0.006)

-0.048*** (0.004)

Border x 2005IPUMS

0.044*** (0.014)

0.007 (0.015)

0.027*** (0.011)

Education

0.026*** (0.001)

0.088*** (0.001)

0.110*** (0.0004)

Education x 2005IPUMS

0.005*** (0.002)

0.015*** (0.003)

0.011*** (0.001)

Experience

0.009*** (0.001)

0.019*** (0.001)

0.027*** (0.0003)

0.003 (0.002)

0.004* (0.002)

0.004*** (0.001)

-0.0001*** (0.00002)

-0.0002*** (0.00002)

-0.0004*** (0.000)

Experience2 x 2005IPUMS

-0.00005 (0.00003)

-0.0001** (0.00005)

-0.0001*** (0.000)

LEP

-0.150*** (0.005)

LEP x 2005IPUMS

-0.030*** (0.010)

Years in U.S.

0.010*** (0.0003)

Years in U.S. x 2005IPUMS

-0.0006 (0.0006) -0.235*** (0.005)

-0.210*** (0.005)

-0.304*** (0.001)

Female x 2005IPUMS

-0.013 (0.009)

0.008 (0.010)

0.022*** (0.002)

2005 IPUMS

0.067** (0.031)

-0.078* (0.043)

-0.014 (0.011)

Constant

1.870*** (0.017)

1.324*** (0.022)

1.074*** (0.007)

R2

.181

.223

.239

N

142,265

107,594

1,738,690

Characteristics

Experience x 2005IPUMS Experience2

Female

Notes: The parentheses contain robust standard errors. These regressions use the IPUMS-provided sampling weights to maintain the national representation of the sample. The symbols ***, **, * denote statistically significant coefficients at the one, five, or ten percent level, respectively. The samples include workers employed for at least 20 hours per week for a minimum of 32 weeks in the previous year, and who resided in a metropolitan statistical area. The equations were estimated with 5 regional dummy variables as additional controls (out of 6 regional dummies: New England, Mid-Atlantic, North Central, South Central, and Mountain), together with their respective interactive terms with 2005 IPUMS. The omitted category is the Pacific region. These 10 additional coefficients are omitted for space constraints.

24

Table 3: Selected Earnings Regression Results for Occupation Dummies (Dependent Variable = Natural Logarithm of Hourly Earnings). Categories and Coefficients in Mexican U.S.-Born Mexican U.S.-Born NonRegressions Immigrants Americans Hispanic Whites Manufacturing workers Border coefficient

-0.122*** (0.014)

-0.161*** (0.017)

-0.025** (0.011)

Border x 2005IPUMS

0.051 (0.034)

0.018 (0.043)

-0.046 (0.032)

Education coefficient

0.022*** (0.001)

0.078*** (0.004)

0.121*** (0.001)

Education x 2005IPUMS

0.006** (0.003)

0.031** (0.008)

0.020*** (0.002)

R2

0.223

0.225

0.308

N

32,376

14,540

267,286

-0.213*** (0.018)

-0.162** (0.022)

-0.045*** (0.015)

Border x 2005IPUMS

0.104** (0.042)

0.098** (0.048)

-0.028 (0.035)

Education coefficient

0.016*** (0.002)

0.049*** (0.005)

0.077*** (0.002)

Education x 2005IPUMS

0.003 (0.004)

0.0009 (0.008)

0.007** (0.003)

R2

0.118

0.141

0.116

N

21,349

8,223

120,789

-0.130*** (0.016)

-0.129*** (0.015)

-0.041*** (0.011)

Border x 2005IPUMS

0.052 (0.036)

0.025 (0.036)

0.022 (0.029)

Education coefficient

0.025*** (0.002)

0.071*** (0.004)

0.105*** (0.001)

Education x 2005IPUMS

0.003 (0.004)

0.006 (0.007)

0.014*** (0.002)

R2

0.172

0.169

0.215

N

18,491

16,609

250,963

Construction workers Border coefficient

Wholesale and retail trade workers Border coefficient

Notes: The parentheses contain robust standard errors. These regressions use the IPUMS-provided sampling weights to maintain the national representation of the sample. The symbols ***, **, * denote statistically significant coefficients at the one, five, or ten percent level, respectively. Other variables in these regressions include those listed in Table 2; the results for these variables can be obtained from the authors. See also notes to Table 2 for an explanation on the regional controls.

25

Page 12: [1] Deleted

amollick

3/7/2007 11:03:00 AM

Second, Table 3 also shows the returns to education in these regressions for the three sectors of activity. In all cases, in agreement to what is expected, construction workers have less favorable impacts to an additional year of education. For Mexican immigrants in the construction sector, for example, the estimated effect of education on earnings is just 0.016 and the returns to education for non-Hispanic whites are generally higher with respect to all categories of workers. Page 12: [2] Deleted

amollick

3/7/2007 11:03:00 AM

Also, I’m not sure if we want to get into the inter-ethnic returns to education within industry here, which is why I think we should delete the second point—that might be a topic that we could explore in a different paper (perhaps not just for the border region, but for the U.S. as a whole…> Page 12: [3] Deleted

3/5/2007 2:35:00 PM

It is interesting to know whether the border penalty has evolved with respect to categories of income Page 12: [4] Deleted

3/5/2007 2:37:00 PM

s. We test whether workers in the low end of the earnings distribution experienced a similar earnings improvement as those at the high end. After estimating e