Fear and lathing in the Michigan furniture industry: employee-based discrimination a century ago.

Buffum, David ; Whaples, Robert

I. INTRODUCTION

Despite progress made in the conceptualization of labor market discrimination, the sources of discrimination continue to elude economists in their empirical investigations (D'Amico [1987]). The primary reason for this failure is inadequate data, since studying the sources of discrimination demands information on workers' institutional settings. In particular, the examination of employee-based discrimination requires information on coworkers' attributes, in addition to the characteristics of the worker himself. Because of this, direct tests of employee-based discrimination in the modern economy are nearly impossible to make. Most tests must be rather indirect.(1)

Here we attempt to examine the sources of discrimination by focusing on discrimination by fellow employees in an era in which intergroup tensions were important, as they are now. We link two themes in labor economics by analyzing employee-based discrimination as a special example of a compensating wage differential. The data, from the 1889 Michigan Bureau of Labor Annual Report on the furniture industry, are uniquely suited to the analysis of employee-based discrimination because they include information on the individual worker, his firm, and his coworkers for over 5000 workers and because the workers surveyed come from several ethnic groups. Using the data, we measure the magnitude and character of employee-based discrimination as we examine specific aspects of ethnic interaction in the labor market in the late 1800s.

Empirically we find that employee-based discrimination was important in this labor market. Workers of one ethnic group received a higher wage when working with members of other groups. Wages rose by about 0.1 percent when the proportion of coworkers in the worker's own ethnic group fell by one percentage point. This response was greater in small cities, small firms, and among certain ethnic groups including the American-born, Dutch, and Poles. Protestants were generally paid more when working with Catholics. Thus, the wage of the furniture worker was determined not only by his ability to "lathe" but also by "fear" and other feelings toward his coworkers.

We close by attempting to measure the additional labor costs generated by employee-based discrimination within a multi-ethnic work force, concluding that they were probably offset by several benefits which rendered complete segregation unnecessary.

II. MODEL AND DATA

Our model of the labor market unites Becker's [1971] model of employee-based discrimination with Rosen's [1974] hedonic wage model. We assume that employers sell goods in a competitive output market and hire mobile workers in a competitive input market. Following the hedonic wage model, firms make job offers to workers that include both wage and nonwage benefits. A worker chooses the wage-nonwage benefit bundle to maximize utility. The nonwage benefits bundle to which we restrict our attention is the ethnic composition of the firm.(2) As the work force of his firm includes more workers whom he likes (dislikes) the worker's wage should fall (rise).

The data from the Seventh Annual Report of the Michigan Bureau of Labor [1890] include the worker's wage, birth place, father's birth place, and several human capital measures: age, occupation, years of experience in occupation, years with present employer, a literacy proxy, place of residence, and marital status for each individual.(3) (Table I shows the characteristics of the American-born, foreign-born, and sons of immigrants in the sample.) The report is organized by firm, with the name and location of the firm followed by a list of the employees in the firm containing thirty-three pieces of information about each individual.(4) Therefore it is possible to measure the ethnic characteristics of each worker's fellow employees, and the effect of work force composition on the worker's wage.

III. EXPECTATIONS ABOUT EMPLOYEE AND EMPLOYER DISCRIMINATION, 1889

Nativism reached a peak in the United States around 1890, and Michigan was a center of nativist feeling. While the most recognizable face of nativism was hostility toward Catholics and the new immigrant groups from Southern and Eastern Europe by the American-born, "often immigrants encountered prejudice and abuse not only from Americans but also from other immigrant groups." "Many of the newcomers distrusted and disliked one another."(5) Although old immigrant groups were not hostile toward each other, there is much evidence that distinctions were strong. For example, Zunz [1982] argues that, within Detroit, class affiliation among workers was secondary to ethnic attachment. Workers were divided along ethnic lines into separate neighborhoods, churches, trade unions, and social clubs.(6)

TABLE I

Characteristics of the Sample

 AMERICAN-BORN SON OF IMM. IMMIGRANT

Percent of total 27.84 20.08 52.08

Wage per day $1.338 $1.156 $1.317

AGE 29.85 22.98 30.25

YEARS IN OCCUPATION 6.45 5.13 7.22

YEARS WITH FIRM 2.65 2.44 2.73

Percent MARRIED 55.65 28.97 56.14

EDUCATION 8.66 7.91 7.57

Percent LITERATE 64.95 45.64 49.68

Percent in GRAND RAPIDS 34.41 49.80 70.10

Percent in DETROIT 3.29 24.01 14.34

Percent in other cities 62.30 26.19 15.56

Percent WITH PARENT 14.66 37.40 20.77

Percent FIRM-BIG 37.41 42.76 59.62

Percent FIRM-MEDIUM 23.25 29.07 19.62

Percent FIRM-SMALL 39.34 28.17 20.76

Percent UNSKILLED 11.65 10.91 16.17

Percent SEMISKILLED 51.79 43.95 36.90

Percent SKILLED 27.54 32.14 39.20

Percent ARTISAN 3.65 8.63 4.78

Percent MANAGERIAL 5.37 4.37 2.95

 Percent born in Percent born in U.S.
 whose parents were born in

Netherlands 22.04 4.34

Germany 14.60 8.87

Scandinavia 4.53 0.14

United Kingdom 3.67 2.60

Poland 2.49 0.42

Canada 2.27 0.46

Ireland 1.32 2.57

Other Catholic 0.82 0.62

Ethnicity was very important in workers' social lives, and the 1889 furniture industry report allows us to assess its economic importance. Accounts of the era suggest two important patterns of ethnic interaction which may be measured with these data.

First, workers may have had a preference for working with members of their own ethnic group rather than with outsiders. Higham's descriptions lead us to expect this of the American-born and we expect this of foreign-born groups, especially the Dutch and Poles, because they are often described as "tight-knit," "isolated," and "insular."(7) There is substantial evidence that this extended directly into the work place. For example, as Ransom [1955, 39] notes, it was generally asserted that the Dutch in Grand Rapids working for the Widdicomb Company would refuse jobs elsewhere at higher pay rather than leave the factory filled with so many countrymen.

Second, workers in specific groups may have preferred not to work with members of other specific groups. Accounts emphasize that natives and old immigrant workers preferred not to work with new immigrants, and that Protestants preferred not working with Catholics.(8)

In historical accounts of this and other eras employee-based discrimination is overshadowed by the emphasis on the foreman and employer as the source of discrimination. Employer-based discrimination may arise in an industry with an uncompetitive output market, and in a labor market in which monopsonistic power and differential supply-elasticities among employee groups are present.(9) However, although we expect to find some evidence that workers are paid more if their foremen are from the same ethnic group, the magnitude of this employer-based discrimination was probably small due to the competitive labor market, especially in Detroit and Grand Rapids (which comprise 70 percent of the sample). Monopsony power was weak: there were many employers; there is no evidence of successful oligopsonistic collusion among the larger employers around 1889; and employment in the industry was rapidly expanding.(10) Likewise, the product market was competitive, for these manufacturers served a national product market which was competitive (Ransom [1955, 22]). While we expect to find evidence of employee-based discrimination, opportunities for employer discrimination were limited. Therefore, we focus on employee-based discrimination.

IV. RESULTS

Wage is assumed to have the functional form lnwage = [B.sub.1][X.sub.1] + [B.sub.2][X.sub.2] + u. Where [X.sub.1] is a vector of worker characteristics, and [X.sub.2] is a vector of coworkers' ethnic characteristics, [B.sub.1] and [B.sub.2] are vectors of coefficients, and u is a random error.

The wage equations in columns 1 and 2 of Table II (which do not include any coworker attributes) explain much of the sample variance, and the magnitude and sign of most of the coefficients are similar to other studies. One surprise is that the wage of single men is significantly higher than that of married men. This higher wage may have been a reward for the greater mobility of single men.(11) Wages are also significantly higher in Grand Rapids. Grand Rapids was noted for its high quality product so this may reflect unmeasured skill differences.(12) The data also include a dummy variable, WITH PARENT, for workers living with their parents and giving them their wages. This lowers the wage by nearly a quarter for a native worker with average traits.(13)

Economists generally measure discrimination as a residual. We estimate a separate OLS lnwage regression for workers in each group. This allows each group to have different coefficients for each productivity-related characteristic, as well as different intercepts. Then we multiply the American coefficients by the average characteristics of the immigrant workers, and subtract the predicted earnings of the immigrant workers using the immigrant workers' own coefficients. This unexplained wage gap is often labeled "discrimination." As given in Table III, this procedure yields an unexplained wage gap of 2.0 to 6.4 percent for immigrants, and an unexplained gap for sons of immigrants which runs from negative 1.8 percent to positive 3.5 percent. The unexplained wage gap has also been estimated for each nationality group separately.

TABLE II

Determinants of the Worker's Wage (Dependent Variable = LN WAGE)

Variable Col. 1 Col. 2

Intercept -130.5(a) (4.1) -120.6(a) (6.9)
SON OF IMM 1.1 (1.3) -6.2(a) (1.7)
IMMIGRANT -3.4(a) (1.1) 4.5 (9.2)
AGE 7.9(a) (0.2) 6.8(a) (0.4)
AGE SQUARED -0.10(a) (0.03) -0.085(a) (0.04)
YEARS IN OCCUPATION 1.3(a) (0.1) 1.0(a) (0.1)
YEARS WITH FIRM 0.9(a) (0.1) 1.1(a) (0.2)
MARRIED -3.8(a) (1.2) -1.3 (2.2)
FIRM-BIG -4.7(a) (1.1) -2.4 (2.0)
FIRM-MEDIUM -0.3 (1.2) -1.3 (2.0)
EDUCATION 0.4(a) (0.2) 0.5(b) (0.3)
LITERATE 7.8(a) (0.9) 6.3(a) (1.8)
GRAND RAPIDS 10.6(a) (1.1) 15.2(a) (1.8)
DETROIT 2.0 (1.5) 9.5(a) (4.3)
SEMISKILLED 10.2(a) (1.2) 11.2(a) (2.3)
SKILLED 14.5(a) (1.3) 15.4(a) (2.6)
ARTISAN 43.1(a) (2.1) 32.5(a) (4.4)
MANAGERIAL 28.0(a) (2.3) 31.2(a) (3.9)
WITH PARENT -27.0(a) (1.2) -20.9(a) (2.4)
S x AGE 5.6(a) (0.7)
S x AGE SQUARED -0.08(a) (0.01)
S x YEARS IN OCCUPATION 0.8(a) (0.3)
S x YEARS WITH FIRM -0.2 (0.4)
S x MARRIED -15.0(a) (3.7)
S x FIRM-BIG -2.9 (3.2)
S x FIRM-MEDIUM -2.0 (3.1)
S x EDUCATION -0.6 (0.5)
S x LITERATE -1.0 (2.7)
S x GRAND RAPIDS -10.1(a) (3.1)
S x DETROIT -11.1(a) (5.0)
S x SEMISKILLED -6.6 (3.7)
S x SKILLED -2.7 (4.0)
S x ARTISAN -2.0 (6.0)
S x MANAGERIAL -9.0 (6.4)
S x WITH PARENT -5.4 (3.4)
I x AGE 0.1 (0.5)
I x AGE SQUARED -0.007 (0.007)
I x YEARS IN OCCUPATION 0.3(a) (0.16)
I x YEARS WITH FIRM -0.3 (0.3)
I x MARRIED 0.8 (2.8)
I x FIRM-BIG -2.3 (2.6)
I x FIRM-MEDIUM 3.7 (2.6)
I x EDUCATION -0.2 (0.3)
I x LITERATE 2.2 (2.2)
I x GRAND RAPIDS -7.0(a) (2.5)
I x DETROIT -10.1(b) (4.7)
I x SEMISKILLED -0.8 (2.9)
I x SKILLED -1.6 (3.0)
I x ARTISAN 19.1(a) (5.4)
I x MANAGERIAL -6.4 (5.2)
I x WITH PARENT -5.4 (3.4)

Number of Observations 5021 5021
Adj R-Squared 64.84 66.2

Notes: I = immigrant; S = son of immigrants. All coefficients have
been multiplied by 100 for ease of presentation. Standard errors are
in parentheses.

a significant at the 99% level. b significant at the 95% level.

These findings are comparable to those of other economic historians.(14) Hannon [1977, 214] drew a one-in-four sample from this same source, used a similar equation, and also found a 6 percent unexplained wage gap for immigrants. McGouldrick and Tannen [1977] found no significant discrimination against Northwestern Europeans in 1890 and 1909. Blau [1980, 32-33] found a 12 percent gap for Northern and Western European immigrants at time of arrival, but the gap disappeared by the time the immigrant was in the country eight or nine years. She concludes that a usable assumption is that old immigrant groups "can provide a standard for nondiscriminatory treatment."(15)

Unexplained wage gaps are small for most groups, nonexistent for many. The wage gap measurement can have a fatal weakness, however. Although the wage gap in this industry indicates that discrimination does little, if anything, to lower the wages of most immigrants, accounts of the period convince us that discrimination was important in determining wages. The wage gap is an aggregate measure and hence is especially weak in assessing discrimination by employees, because several groups may discriminate against each other and the effects this has on wages may cancel out. We solve this problem by introducing firm-level ethnic composition measures into the individual's wage equation.

Theoretically, there are several possible ways in which wages will change as co-worker ethnicity changes. A simple assumption is that the wage will rise at an unchanging rate as the percent of outsiders rises and the percent of own group, therefore, falls.(16) Group definition is somewhat ambiguous, however. For example, another individual can be considered a member of your group if (A) he is born in the same country as you, (B) his parents were born in the same country as your parents, or (C) he was born in the same country as you and his parents were born in the same country as your parents. Definition A assumes rapid assimilation and puts sons of immigrants of all nationalities in the same group along with the American-born, while removing sons of immigrants from their parents' group. Definition B assumes slow assimilation and puts sons of immigrants in the same group as their parents. Definition C is narrower, putting sons of immigrants of each ethnicity in different groups, separate from their parents and the American-born.

TABLE III

Unexplained Wage Gaps (in percent)

 Col. 1 Col. 2

All Immigrants 6.4 2.0

Sons of Immigrants 1.8 -3.5

Immigrants from: Ireland -7.2 -7.9

 United Kingdom -1.8 -2.1

 Canada 1.4 3.8

 Scandinavia 1.9 -0.2

 Germany 4.7 3.0

 Other Catholic 4.8 1.2

 Netherlands 8.9 -2.2

 Poland 18.1 -2.1

[WAGE.sub.N] - [WAGE.sub.I] = [B.sub.N][X.sub.N] -
[B.sub.I][X.sub.I] where, WAGE = the predicted ln wage, B = vector
of coefficients in Table II, column 1, X = vector of average
characteristics, N = American-born, I = Immigrant. Therefore,
[WAGE.sub.N] - [WAGE.sub.I] can be expressed as (1) or (2) below,

(1) [WAGE.sub.N] - [WAGE.sub.I] =
[B.sub.N]([X.sub.N]-[X.sub.I])+([B.sub.N]-[B.sub.I])[X.sub.I]

(2) [WAGE.sub.N] - [WAGE.sub.I] =
[B.sub.I]([X.sub.N]-[X.sub.I])+([B.sub.N] - [B.sub.I])[X.sub.N]

where the first term on the right-hand side is the explained wage
gap, and the second term is the unexplained wage gap. The
unexplained wage gap in (1) is given in column 1, and the
unexplained wage gap in (2) is given in column 2.

Table IV presents wage-(percent own group) coefficients using definitions A (% OWN GROUP), and B (% PARENTS' GROUP) using the same variables as in column 2 of Table II and the linear form of % OWN GROUP and % PARENTS' GROUP. The coefficients are statistically [TABULAR DATA FOR TABLE IVA OMITTED] significant and have the predicted sign. A one percentage point increase in the percentage of a worker's own group in the work place lowers wages by about 0.11 to 0.12 percent. A one percentage point increase in the percentage of a worker's parents' group in the work place lowers wages by about 0.07 to 0.09 percent. Members of the same ethnic group are probably able to communicate better with one another than with outsiders. This communication effect will raise a worker's productivity, and thus his wage, as he works with more members of his own ethnic group. The negative wage-ethnic composition coefficients we find are therefore even stronger evidence of the importance of worker preferences, since they overwhelm this opposing communication effect. In Table IV, we also estimate these coefficients for each ethnic group. A more negative coefficient should indicate a group that is more hostile to outsiders and/or has stronger internal bonds. As the historical literature suggests, the Dutch, Polish, and American coefficients are greater in magnitude than the coefficient for all workers together.

TABLE IVB

Coefficients on % PARENTS' GROUP by Nationality (Dependent variable
= LN WAGE)

Polish -0.880
 (0.701)

German -0.016
 (0.069)

Irish -0.220
 (0.414)

Dutch -0.151(a)
 (0.064)

U.S. -0.172(a)
 (0.046)

Notes: The regression controls for the variables in Table II, column
1. Standard errors are in parentheses.

a significant at the 99% level.
TABLE V

Coefficients on % OWN GROUP by Decile (Dependent variable = LN WAGE)

% OWN GROUP

No others 10.8(a)
[1,10] 1.8
[11,20] -2.1
[21,30] Control
[31,40] -1.6
[41,50] -4.2(b)
[51,60] -4.8(b)
[61,70] -3.7
[71,80] -4.6
[81,90] -9.1(a)
[91,99] -9.3(a)
100 none

Note: These coefficients result from a regression of the type in
column 2 of Table II.

a significant at the 99% level.

b significant at the 95% level.

A less restrictive way to assess the wage-ethnic composition relationship uses a set of dummies measuring the percentage of a worker's own group in the work place. A representative set of dummies is shown in Table V. An individual whose group makes up about one-quarter of the firm receives a substantially lower wage than one who is the sole member of his ethnic group in the firm, about 11 percent lower. He will also give up an additional 9 percent of his wage to work at a firm composed almost entirely of his own group, and about 5 percent of his wage to work where his group makes up about half the work force.

The data allow us to simultaneously test for one type of employer discrimination. The foreman-worker ethnic interaction is captured in two ways: FOREMAN (PARENT) is a dummy variable for the presence in the firm of a foreman from the worker's (parents') ethnic group; FOREMANHIGH (PARENT) is a dummy equaling one if the percentage of the foremen of the worker's (parents') ethnic group is greater than the percentage of that ethnic group in the firm. The coefficients on FOREMAN indicate that workers actually received less when their firm had a foreman of their nationality. This may be a better measure of who isn't the boss, than of who is the boss, however. The variable FOREMANHIGH tells the expected story that the wage rises (by about 2 to 3 percent) when there is a surplus of foremen from the worker's group. These magnitudes are not large compared to the effect of % OWN GROUP. Thus, the emphasis which some historians put on foremen as the root of discrimination in this era seems to be overstated.(17)

The coefficients on the percentage of a worker's own group in the workforce (% OWN GROUP) and the foreman dummies paint a general picture of ethnic relations in the Michigan furniture industry, but the data also let us see what effect the institutional setting had on the operation of discrimination. In Table VI, we present % OWN GROUP and foreman coefficients separately for Grand Rapids, Detroit, and the smaller towns. The coefficients % OWN GROUP are larger in magnitude in the small cities. This supports Hannon's [1982] argument that immigrants found the climate of discrimination milder in larger cities. Table VI also gives % OWN GROUP and foreman coefficients by firm size. In the large firms the % OWN GROUP coefficient is insignificant. The problem (for workers and firms) of hiring workers from different groups probably disappeared in larger factories because any segregation could be carried out in-house.(18) The foreman coefficients do not vary greatly by firm size.

Another important institutional feature of the late 19th-century factory is the rapid turnover of labor (Jacoby and Sharma [1992]). The average tenure of the employees in the firm may be an important influence on discrimination. One might expect that employee-based discrimination is more important in firms where turnover is low and the workers stay together as a group for a long time. However, the % OWN GROUP coefficient does not seem to be affected by the stability of the work force. The coefficient is -0.099 for workers in firms with below-median employer tenure, and -0.105 for those above the median.

In Table VII we examine ethnic interaction within occupations within firms. The variable % OWN IN OCC refers to the percent of the employees in the worker's firm and occupation who are of the same ethnic group as he. Becker argues that workers will be unable to show their prejudice against coworkers of other groups with whom they are close substitutes, despite holding the most resentment toward them. We find this is generally true. The interaction coefficients within occupations are smaller in magnitude than within the entire firm, are never significant, and only approach significance in smaller firms. In part 2 of Table VII, we report the coefficients on % OWN IN OCC for regressions restricted by occupation. Coefficients are generally small and insignificantly different from zero, but the few significant coefficients are surprisingly in unskilled occupations, where substitutes should be easily found.

TABLE VI

Coefficients on % OWN GROUP, % PARENTS' GROUP and FOREMAN by City
and Firm Size (Dependent variable = LN WAGE)

 Grand Rapids Detroit Smaller Towns

% OWN GROUP -0.032 -0.061 -0.214(a)
 to -0.087(b) to -0.103(c) to -0.245(a)
% PARENTS' GROUP -0.043 -0.001 -0.181(a)
 to -0.089(a) to -0.031 to -0.190(a)
FOREMAN -3.7(a) 4.4 -0.5
FOREMAN (PARENT) -2.7(b) -3.3 -2.5
FOREMANHIGH 2.7(b) -6.4(b) 0.6
FOREMANHIGH (PARENT) 4.2(a) 20.9 -3.7(b)

 Small Medium Large

% OWN GROUP -0.183(a) -0.129(a) +0.043
 to -0.234(a) to -0.156(b) to -0.017
% PARENTS' GROUP -0.135(a) -0.068(c) +0.007
 to -0.192(a) to -0.075(b) to -0.036
FOREMAN -5.1(b) 2.5 -4.3(a)
FOREMAN (PARENT) -6.0(b) -0.1 -3.2(b)
FOREMANHIGH 3.6 3.3 2.6(b)
FOREMANHIGH (PARENT) 3.2 1.4 1.6

Notes: Each column of coefficients is derived from four regressions
of the type given in Table IVA. The regressions control for the
variables in Table II, column 2.

a significant at the 99% level.

b significant at the 95% level.

c significant at the 90% level.

Another reason to look below the aggregate level is that the coefficient on % OWN GROUP compounds two effects the worker's preference for working with members of his own group and his aversion to working with other groups. A richer picture of ethnic relations can be drawn by estimating log of wage equations for each ethnic group separately (U.S., German, Dutch, Irish) and including the percent of each specific ethnic group in each equation.(19) Table VIII presents these ethnic interaction coefficients. Although [TABULAR DATA FOR TABLE VII OMITTED] the picture is not exceptionally clear, our expectations about patterns of nativism are generally fulfilled. Most significant coefficients are positive, thus workers received more for working with members of other ethnic groups. Several relatively large positive coefficients implying hostility stand out, however. The primarily Protestant Americans were most averse to working with immigrant Catholics (i.e., Irish, Polish, Canadian, and other Catholics), but not averse to working with Protestant immigrants (German, Dutch, British, and Scandinavian). This accords well with the anti-Catholic nature of nativism of this era.(20) German immigrants also received a premium for working with Catholic groups (Irish, Canadians, and other Catholics). The largest premium of the Calvinistic Dutch immigrants was for working with the Irish. (However, the premium is negative for Dutch immigrants working with Poles.) Finally, the Irish were averse to working with Scandinavians.

TABLE VIII
Ethnic Interaction Coefficients

Read each column down to find how much an additional percentage
point of the group listed on the side affects the wage of the group
whose parent's ethnicity is listed at the top of the column. In part
A the group listed on the side is measured as the percent of workers
whose parents were born in the location. In part B the group listed
on the side is measured as the percent of workers born in the
location.

 U.S. Netherlands Germany Ireland

Percent of workers whose PARENTS were born in:

U.S. 0.2 -0.1 -0.1
Netherlands 0.1 0.0 -0.2
Germany 0.1 0.4(a) 0.0
Ireland 0.1 0.3 0.7(a)
United Kingdom 0.2 0.4 0.0 -1.0
Scandinavia 0.2 -0.1 0.1 1.1(b)
Poland 0.9(a) -1.0(a) -0.4 -0.2
Canada 0.5(b) 0.4 1.5(a) 0.4
Other Catholic 1.5(a) -1.9(b) 0.9(a) 0.7

Percent of WORKERS who were born in:

U.S. 0.2(b) 0.0 -0.1
Netherlands 0.1 0.0 -0.1
Germany 0.1 0.3 -0.4
Ireland 1.3(a) 1.6(a) 1.4(b)
United Kingdom 0.5 1.0 0.3 -1.7
Scandinavia 0.2 0.4 0.2 0.0
Poland 1.1(a) -1.0(b) -0.2 0.0
Canada 0.3 -0.5 0.9(a) 0.4
Other Catholic 3.0(a) 0.6 0.2 1.0

Note: These coefficients result from a regression of the type in
Table II, column 1.

a significant at the 99% level.

b significant at the 95% level.

V. EMPLOYEE DISCRIMINATION AND SEGREGATION

Models of employee-based discrimination generally imply that the work force should be segregated along ethnic lines (Becker [1971]). We have found evidence of employee discrimination, yet this industry is not completely segregated by ethnic group. The findings help resolve this paradox.

The difference in cost between a non-segregated work place and a segregated work place can be estimated using these data. We measure the additional cost of a nonsegregated firm three ways; one simple, two more complex. All yield similar estimates. The additional cost can be approximated by (100 - the mean % OWN GROUP) x (the coefficient of % OWN GROUP), since in a segregated firm the mean of % OWN GROUP is 100. The estimated additional labor cost for non-segregation in the Michigan furniture industry, according to this estimate, is about 6 percent. (7.2 percent in Detroit, 5.2 percent in Grand Rapids, and 6.0 percent in the smaller cities.)

The second and third estimates assume two types of firms, A and B. Firm A's work force consists only of workers with the mean of native traits, whose wages are determined by American coefficients, and which is composed of workers of only one ethnic group. Firm [B.sub.i]'s workers have the same traits and wage equations as firm A's, but firm [B.sub.i] has the same percent of workers from each ethnic group that existed in the ith firm in the Michigan furniture industry. The cost to firm i of nonsegregation is estimated as the sum of

(100 - % OWN [GROUP.sub.j]) x (% OWN [GROUP.sub.j]) x (% OWN GROUP coefficient),

where j = 1, ..., N are each of the ethnic groups. The cost of nonsegregation for the median firm is 6.8 percent.

The cost advantage of segregated firm A can also be estimated by using the ethnic interaction coefficients in Table VIII. This yields an estimate which is virtually the same as the first two. The median firm's wage bill is 6.1 percent higher than firm A's. Since wages are only about 36 percent of the wholesale value of output among these firms, the cost of nonsegregation may not be overwhelming.

However, even a small increase in the firm's wage bill should generate segregation in most models which assume competition. Why wasn't there complete segregation in this market? Probably because the assumptions of these models are too extreme. They assume that ethnic composition enters the firm's profit function only by generating compensating wage differentials. In fact, a multi-ethnic work force probably yields some benefits which reduce costs, offsetting the increased costs generated by employee-based discrimination.

As Arrow [1973] emphasizes, workers of different ethnic groups are not perfect substitutes. Ethnic mixing probably reduced firms' (and workers') search costs. In the late 1800s firms did little to actively search for workers. The two most widely used employer "search" strategies were interviewing neighborhood workers who walked in off the street with little or no information about openings, and using current employees to spread news of openings by word of mouth (De Schweinitz [1932], Jacoby [1985]). Licht's [1992] study of Philadelphia workers who entered the labor force between 1872 and 1904 shows that very few found their jobs through institutional searches, such as advertisements or employment agencies. Fifty-nine percent found intermediate jobs through "personal initiative" (predominately walking from factory to factory). About half as many lined up the job via "connections" (mostly friends). This type of search probably yielded a mix of applicants reflective of the ethnic composition in the surrounding neighborhood.(21) To generate a pool of applicants across all skill levels that consisted only of one ethnic group entailed considerably higher search costs.(22) In this era before the automobile the average journey to work was very short, allowing the firm to attract few workers from more than a mile away.(23) The median completed job spell among these workers was only two years. This high employee turn-over also made institutional searches by the employer expensive during this period, since the investment had to be recouped over such a short period.(24)

While ethnic mixing reduced short-run search costs, long-run wages may have also been lowered since ethnic mixing reduced collective action and solidarity among workers.(25) Segregation may not have been viable for other long-term reasons. The ethnic composition of the labor force in Michigan and in this industry changed drastically after 1890 as the percent of employees who were Polish grew rapidly, and the Dutch, German, and Irish shares fell.(26) A firm which used the short-run expedient of an all-Dutch work force would have been faced with a rapidly declining supply of workers. Moreover, the employee interaction coefficients were not chiseled in stone. The Americanization of the work force probably eroded discrimination. Furthermore, as Higham argues, ethnic hostility was peaking at this time, so the long-run coefficients were likely to have been lower. These factors would deter an employer from getting locked into a segregated work force.

Finally, as Atack [1985] demonstrates, the efficient scale of plants was increasing rapidly in the late 1800s. If a firm planned on moving to the larger optimal size it did not need to adopt a short-run solution to the problem of ethnic hostility (segregation of the firm) when the long-run solution (segregation within the firm) was opposed to it. These benefits of nonsegregation could easily have outweighed the relatively small costs of bringing discriminating workers together.

VI. CONCLUSIONS AND IMPLICATIONS

Only a few economists have examined the sources of discrimination. Unfortunately, data limitations mean this search often relies on indirect evidence.(27) In this study we look directly at discrimination, using data from the late 1800s to derive wage-ethnic composition coefficients. We confirm the employee-based discrimination hypothesis, showing that the ethnic composition of the work place did significantly affect a worker's wage. The wage-ethnic composition coefficients are roughly two-thirds the size of similar co-efficients found a century later by Ragan and Tremblay [1988] for young employees whose coworkers are of a different race.

We also show where employee-based discrimination thrived best: in small towns and small firms. We broaden the view of employee discrimination by showing that minorities, as well as the majority, discriminated. The picture of discrimination is not necessarily malicious, however, since much of it seems to be due to camaraderie among members of the same group. Finally we attempt to measure the cost of segregation, and show that the cost to the firm of a multi-ethnic work force was probably too low in 1889 to have fostered total segregation.

1. The sole direct test made with modern data uses self-reported data on the racial attributes of coworkers among young employees in the National Longitudinal Survey (Ragan and Tremblay [1988]). Its results are similar to ours. Workers of each race are compensated more when working with members of other races. However, the measures of coworker attributes are much simpler than ours, and differences by institutional setting cannot be examined. In a less direct test, Chiswick [1973] examined the variance in wages, finding that whites' wages vary more when there are more blacks in the labor market. Although this evidence supports the employee-based discrimination hypothesis, the test cannot tell the magnitude of this discrimination.

2. There are many other nonwage benefits which may be of importance to the worker, especially unemployment risk. We do not believe that worker preference for these other benefits is systematically related to ethnic composition, and have estimated the determinants of days employed, finding no significant ethnic component.

3. Ransom and Sutch [1991] describe the data. About 4 percent of the workers were female. For consistency they were omitted from the analysis.

4. Most firms provided information on the size of their work force. In these firms 59 percent of the workers were included in the annual report. The report is unclear about which employees were canvassed and which were left out. There is no way of knowing the biases in the collected data, but the authors of the Michigan Bureau of Labor and Industrial Statistics's report [1890, xi] believed they were not important. "While the canvas does not include the full force employed in the industry, enough has been obtained to show a good average, and the conditions shown are a fair representation of the conditions surrounding those employed in this industry throughout the state."

5. Higham [1981] identifies four peaks in American nativism: the late 1790s, the 1850s, 1886 to 1896, and the period during and after World War I. The greatest strength of the American Protective Association, the leading nativist anti-Catholic organization of the 1890 period, was in Michigan. The quotes are from Matulich [1971, 110], and Dinnerstein and Reimers [1988, 65].

6. Zunz [1982, 3-6, 13]. Likewise, Kleiman [1985, ix], argues that in Grand Rapids "workers ... centered their lives on home-ownership, church and ethnic groups." Examples of organized labor divided along ethnic lines abound. In Grand Rapids, for example, a German Furniture Workers Association existed, (Ransom [1955, 42]), while Detroit had a German Central Labor Union (Orton [1981, 179]). In Chicago the Furniture Workers Union had six separate organizations for six different language groups (Darling [1984, 53]). Whaples and Buffum [1991] find significant links between ethnic concentration and the purchase of insurance among furniture workers. Bernard [1980] shows that German, Irish, Polish, and Scandinavian immigrants in Wisconsin rarely married outside their ethnic group.

7. See for example, Orton [1981, 170], Vanderstel [1983, 70-71], and Kleiman [1985, 39], who argues that "the insularity of Grand Rapids' Hollanders can be traced to their desire for a pure, reformed church.... [F]rom the very earliest settlement, the Dutch had pursued deliberate cultural and ecclesiastical isolation."

8. Higham [1981] emphasizes anti-Catholicism. Kleiman [1985, 42-43, 67], stresses the "deep-seated ethnic and religious differences among residents" of Grand Rapids, and gives the example of German-Polish tension. "The Germans ... never really let the Poles forget that ... their native country had been absorbed into the German empire." See also, Matulich [1981, 111], Dinnerstein and Reimers [1988], Brault [1986], and Carpenter [1970].

9. Becker [1971, 39-54, 110] and Madden [1977].

10. In 1890, Detroit had 16 furniture factories and 1746 manufacturing establishments, Grand Rapids had 31 and 869. Monopsony power was weak because every furniture factory in Grand Rapids and Detroit was within a mile of another, most were within a half mile of another (Polk [1891] and Kolts [1891]). Furniture industry employment grew 117 percent in Detroit and 142 percent in Grand Rapids between 1880 and 1890, outstripping population growth by 29 percent and 55 percent respectively. (United States Census Office [1883, 399, 403] and United States Census Office [1895, 194, 230]).

The Furniture Manufacturers Association of Grand Rapids was established to pool resources in obtaining lower transportation and insurance rates, not to monopsonize the labor market. Because the employers did not show a united front against the workers' eight-hours strikes of May 1886, the organization virtually ceased functioning for several years (Ransom [1955, 46-47]). Later, however, labor market monopsony power may have arisen. Around 1905, however, labor market monopsony power may have arisen, when some of the Grand Rapids furniture manufacturers created an Employers Association (Kleiman [1985, 78]).

11. Korenman and Neumark [1991] show the universality of positive marriage premia in modern studies. The internal labor markets of today do not reward job shifting as much as the spot labor market of the late 1800s. Lower employer tenure and greater layoff rates are found for single men, when all else is controlled, in this sample. Whaples [1991] finds evidence of negative marriage premia in several industries and locations in the late 1800s.

12. According to Ransom [1955, 63), Grand Rapids furniture had such a good name that out-of-state manufacturers fraudulently labeled their furniture, "Made in Grand Rapids."

13. This is probably because young workers who lived with their parents had lower living expenses and were able to forego initial earnings while investing in on-the-job training and/or because supplying labor as a family member rather than as an individual constrained job choice and led to a less perfect matching of worker and job and/or because workers who earned less may not have been able to afford to live away from home. The endogeneity of the place-of-residence decision is examined in Whaples [1992]. The omission or inclusion of the WITH PARENT variable does not affect any of the results concerning the source of discrimination.

14. Following our lead, Craig and Fearn [1993] have discovered compensating wage-differentials among white American and Northern European seamen in the whaling industry in the 1850s.

15. The generally cited weakness of the wage-gap approach is that it implicitly assumes that none of the gap is due to unmeasured productivity differences between the groups. One unmeasured characteristic which is related to productivity is the ability to speak English. Its omission will bias the unexplained gap upward. Higgs [1971] using aggregate data from around 1910 found that literacy and the ability to speak English explained almost four-fifths of the variance in average earnings among ethnic groups.

16. Becker [1971, 55-74], for example, assumes that working with a member of the other group is a zero-one dummy. Blalock [1956], on the other hand, emphasizes the importance of the size and changes in the size of the other group. Antos and Rosen [1975] have calculated the influence on teachers' wages of the proportion of the student body that is white. Our measure parallels that of Antos and Rosen.

17. Jacoby [1985, 17] emphasizes the foreman's power in hiring, firing, and setting wages, and his role in ethnic discrimination. His description of foreman power and caprice may be overstated for the Michigan furniture industry (and the furniture industry in general) where there was a tradition of close owner-worker contact. As Ransom [1955, 47] and Darling [1984, 66] report, many of the owners in this industry had climbed from initial positions as workers. However, Nelson [1975] notes that foreman discretion increased for "batch job" work, which characterized many of these firms. Thus we are unable to gauge the representativeness of this result. In any case, we are unable to determine who works with each foreman, so we urge caution in interpreting the foreman-worker interactions.

18. Whatley [1990, 56] provides examples of in-house racial segregation along occupational lines.

19. We also estimated these coefficients for Polish, Canadian, British, and Scandinavian workers, but they were all statistically insignificant due to small sample sizes.

20. The variable Other Catholic consists mostly of French and Belgian, with a few Swiss immigrants. Approximately one-fifth of the Canadians in Michigan were French Canadians, and one-third of the Canadians in our sample had parents born in Ireland.

21. Neighborhoods were not so segregated that this would yield applicants of only one ethnic group. Instead, neighborhoods were often dominated by one ethnic group, but with "pockets" of several other ethnic groups intermixed. None of Detroit's or Grand Rapids' neighborhood approached complete ethnic segregation (Zunz [1982], Vanderstel [1983]).

22. Wright [1986] argues that all-Black cotton mills in the postbellum South failed because search costs for skilled black workers were too high (almost infinite). Likewise, the search costs for skilled Polish furniture workers would have been extremely high, making a segregated all-Polish furniture factory unprofitably expensive.

23. Hershberg et al. [1981], using data from Philadelphia in 1880, show that the median journey to work was .5 miles for carpenters, .58 miles for cabinet makers, and .51 miles for iron factory workers. These distances are probably a bit lower than in Detroit or Grand Rapids since Philadelphia was more densely settled.

24. If the search cost per worker rose as the number of workers rose, then the costs of segregation were higher in larger firms, and smaller firms should have been more segregated. The Herfindahl-Hirschman index of ethnic group concentration is lowest for firms with over seventy-five employees (.353, using the PARENT group definition, .372, using the OWN group definition), higher for firms with twenty-six to seventy-five employees (.479, .495), and highest for very small firms with twenty-five or fewer workers (.535, .557).

25. William Jones, Superintendent of Andrew Carnegie's massive Edgar Thompson steel works expressed this very idea. "My experience has shown that Germans and Irish, Swedes, and what I denominate "Buckwheats," young American country boys, judiciously mixed, make the most ... tractable force you can find" (Bridge [1903, 81]). The cost of unionization could be high. Mullin [1993] argues that typical union premiums in this period were about 14 percent. Eichengreen [1987] estimates that the union wage premium in Iowa in 1894 was 34 percent. Radical labor leader William Haywood in an editorial entitled "What's the Matter with Grand Rapids?" cited ethnic divisions among immigrant groups as one of the critical factors inhibiting the advance of class-conscious struggle and unionization there. Ethnic division among the Dutch and Poles was central to the failure of the city-wide strike by furniture workers in 1911 (Kleiman [1985, 35, and chapter 4]).

26. Ransom [1955, 64] notes the influx of Poles into the Grand Rapids furniture industry by World War I. The percent of all immigrants, and especially of the Irish, German, and Dutch, in Kent (Grand Rapids) and Wayne (Detroit) counties fell from 1890 to 1920, while the supply of Eastern Europeans grew. The following table gives the percent of the population in 1890 and 1920 made up of immigrants from these countries.

27. For example, Bergmann and Lyle [1971] and Gwartney and McCaffree [1971] both examine employee-based discrimination through evidence of occupational crowding. Borjas and Bronars [1989] examine evidence on self-employment to assess customer-based discrimination.

 Kent Wayne
 1890 1920 1890 1920

All Immigrants 49.67 18.85 46.64 29.27
Dutch 20.37 7.92 0.18 0.17
German 7.08 1.59 21.08 2.95
Irish 3.67 0.43 4.11 0.68
Polish 1.95 2.82 2.72 6.43
Other Eastern
Europe 0.56 1.67 0.99 5.58

Sources: U.S. Census Office [1892], and U.S. Bureau of the Census
[1922].

REFERENCES

Antos, Joseph, and Sherwin Rosen. "Discrimination in the Market for Public School Teachers." Journal of Econometrics, May 1975, 123-50.

Arrow, Kenneth. "The Theory of Discrimination," in Discrimination in Labor Markets, edited by Orley Ashenfelter and Albert Rees. Princeton: Princeton University Press, 1973, 3-33.

Atack, Jeremy. "Industrial Structure and the Emergence of the Modern Industrial Corporation." Explorations in Economic History, March 1985, 29-52.

Becker, Gary. The Economics of Discrimination, 2nd. ed. Chicago: University of Chicago Press, 1971.

Bergmann, Barbara, and Jerolyn Lyle. "The Occupational Standing of Negroes by Area and Industry." Journal of Human Resources, Fall 1971, 411-33.

Bernard, Richard. The Melting Pot and the Altar: Marital Assimilation in Early Twentieth Century Wisconsin. Minneapolis: University of Minnesota Press, 1980.

Blalock, H. M. "Economic Discrimination and Negro Increase." American Sociological Review, October 1956, 584-88.

Blau, Francine. "Immigration and Labor Earnings in Early Twentieth Century America." Research in Population Economic, 2, 1980, 21-41.

Borjas, George, and Stephen Bronars. "Consumer Discrimination and Self-Employment." Journal of Political Economy, June 1989, 581-605.

Brault, Gerard. The French-Canadian Heritage in New England. Hanover, N.H.: University Press of New England, 1986.

Bridge, James. The Inside History of the Carnegie Steel Company. New York: Aldine, 1903.

Carpenter, Niles. Nationality, Color and Economic Opportunity in the City of Buffalo. Westport, Conn.: Negro Universities Press, 1970.

Chiswick, Barry. "Radal Discrimination in the Labor Market: A Test of Alternative Hypotheses." Journal of Political Economy, November 1973, 1330-52.

Craig, Lee, and Robert Fearn. "Wage Discrimination and Occupational Crowding in a Competitive Industry: Evidence from the American Whaling Industry." Journal of Economic History, March 1993, 123-38.

D'Amico, Thomas. "The Conceit of Labor Market Discrimination." American Economic Association Papers and Proceedings, May 1987, 310-15.

Darling, Sharon. Chicago Furniture: Art, Craft, and Industry, 1833-1983. New York: Norton, 1984.

De Schweinitz, Dorothea. How Workers Find Jobs: A Study of Four Thousand Hosiery Workers in Philadelphia. Philadelphia: University of Pennsylvania Press, 1932.

Dinnerstein, Leonard, and David Reimers. Ethnic Americans: A History of Immigration and Assimilation. New York: Harper Collins, 1988.

Eichengreen, Barry. "The Impact of Late Nineteenth-Century Unions on Labor Earnings and Hours." Industrial and Labor Relations Review, July 1987, 501-15.

Gwartney, James, and Kenneth McCaffree. "Variance in Discrimination among Occupations." Southern Economics Journal, October 1971, 141-55.

Hannon, Joan. "The Immigrant in the Promised Land: Human Capital and Ethnic Discrimination in the Michigan Labor Market, 1888-1890." Ph.D. dissertation, University of Wisconsin, 1977.

-----. "City Size and Ethnic Discrimination: Michigan Agricultural Implements and Iron Working Industries, 1890." Journal of Economic History, December 1982, 825-45.

Hershberg, Theodore, Harold B. Cox, Dale B. Light, Jr., and Richard R. Grenfield. "The 'Journey to Work:' An Empirical Investigation of Work, Residence and Transportation, Philadelphia, 1850 and 1880," in Philadelphia: Work, Space, Family and Group Experience in the Nineteenth Century, edited by Theodore Hershberg. New York: Oxford University Press, 1981, 128-73.

Higgs, Robert. "Race, Skills, and Earnings: American Immigrants in 1909." Journal of Economic History, June 1971, 420-28.

Higham, John. Strangers in the Land: Nativism in American History, 1860-1925. New York: Atheneum, 1981.

Jacoby, Sanford. Employing Bureaucracy: Managers, Unions, and the Transformation of Work in American Industry, 1900-1945. New York: Columbia University Press, 1985.

Jacoby, Sanford, and Sunil Sharma. "Employment Duration and Industrial Labor Mobility in the United States, 1880-1980." Journal of Economic History, March 1992, 161-79.

Kleiman, Jeffrey. "The Great Strike, Religion, Labor and Reform in Grand Rapids, Michigan, 1890-1916." Ph.D. dissertation, Michigan State University, 1985.

Kolts, Matthew J., and Sons. City Guide and Directory of Grand Rapids. Grand Rapids, 1891.

Korenman, Sanders, and David Neumark. "Does Marriage Really Make Men More Productive?" Journal of Human Resources, Spring 1991, 282-307.

Licht, Walter. Getting Work: Philadelphia, 1840-1950. Cambridge, Mass.: Harvard University Press, 1992.

Madden, Janice. "A Spatial Theory of Sex Discrimination." Journal of Regional Science, December 1977, 369-80.

Matulich, Loretta. "A Cross-Disciplinary Study of the European Immigrants of 1870-1925." Ph.D. dissertation, University of New Mexico, 1971.

McGouldrick, Paul, and Michael Tannen. "Did American Manufacturers Discriminate against Immigrants Before 1914?" Journal of Economic History, September 1977, 723-46.

Michigan Bureau of Labor and Industrial Statistics. Seventh Annual Report. Lansing, 1890.

Mullin, Debbie Dudley. "The Porous Umbrella of the AFL: Evidence from Late Nineteenth-Century State Labor Bureau Reports on the Establishment of American Unions." Ph.D. dissertation, University of Virginia, 1993.

Nelson, Daniel. Managers and Workers: Origins of the New Factory System in the United States, 1880-1920. Madison: University of Wisconsin Press, 1975.

Orton, Lawrence. Polish Detroit and the Kolasinski Affair. Detroit: Wayne State University Press, 1981.

PoLk, R. L., and Company. Detroit City Directory. Detroit, 1891.

Ragan, James, and Carol Tremblay. "Testing for Employee Discrimination by Race and Sex." Journal of Human Resources, Winter 1988, 123-37.

Ransom, Frank. The City Built of Wood: A History of the Furniture Industry in Grand Rapids, Michigan, 1850-1950. Ann Arbor: Edwards Brothers, 1955.

Ransom, Roger, and Richard Sutch. Codebook and User's Manual: A Survey of 5419 Workers in the Furniture Industry of Michigan, 1889. Berkeley: Institute of Business and Economic Research, University of California, 1991.

Rosen, Sherwin. "Hedonic Prices and Implicit Markets." Journal of Political Economy, January 1974, 34-55.

United States Census Office. Tenth Census of the United States, 1880, Volume 2, Manufactures. Government Printing Office: Washington, D.C., 1883.

-----. Eleventh Census of the United States, 1890, Volume 1, part 1, Population. Government Printing Office: Washington, D.C., 1892.

-----. Eleventh Census of the United States, 1890, Volume 6, part 1, Statistics of Cities, Manufacturing Industries. Government Printing Office: Washington, D.C., 1895.

-----. Fourteenth Census of the United States, Volume III, Population: 1920. Government Printing Office: Washington, D.C., 1922.

Vanderstel, David. "The Dutch in Grand Rapids, Michigan: A Study of Social Mobility in a Midwestern Urban Community, 1850-70." Ph.D. dissertation, Kent State University, 1983.

Whaples, Robert. "The Marriage Premium in Historical Perspective." Paper presented at the National Bureau of Economic Research, Conference on Historical State Labor Statistics, Lawrence, Kansas, June 1991.

Whaples, Robert. "Age at Leaving Home in the Late 1800s: An Economic View Through the Eyes of the Young." Paper presented at the Social Science History Association meetings, Chicago, November 1992.

Whaples, Robert, and David Buffum. "Fraternalism, Paternalism, the Family and the Market: Insurance a Century Ago." Social Science History, Spring 1991, 97-122.

Whatley, Warren. "Getting a Foot in the Door: 'Learning,' State Dependence, and the Racial Integration of Firms." Journal of Economic History, March 1990, 43-66.

Wright, Gavin. Old South, New South: Revolutions in the Southern Economy since the Civil War. New York: Basic Books, 1986.

Zunz, Olivier. The Changing Face of Inequality: Urbanization, Industrial Development, and Immigration in Detroit, 1880-1920. Chicago: University of Chicago Press, 1982.

APPENDIX

Definitions of Variables

LNWAGE = natural log of worker's daily wage in dollars

SON OF IMM = 1 if parent was not born in the United States; 0 otherwise (SON OF IMM is abbreviated S when it is used interactively.)

IMMIGRANT = 1 if worker was not born in the United States; 0 otherwise (IMMIGRANT is abbreviated I when it is used interactively.)

YEARS IN OCC = years in present occupation

YEARS WITH FIRM = years with current employer

MARRIED = 1 if worker is married or widowed; 0 otherwise

FIRM-BIG = 1 if worker's firm has more than 150 employees canvassed in the survey; 0 otherwise. Fourteen of the seventy-eight firms are big.

FIRM-MEDIUM = 1 if worker's firm has more than 75 employees and less than 150 employees canvassed in the survey; 0 otherwise. Seventeen of the seventy-eight firms are medium.

EDUCATION = age began work - 6

LITERATE = 1 if worker receives a newspaper or magazine; 0 otherwise

GRAND RAPIDS = 1 if worker works in Grand Rapids; 0 otherwise

DETROIT = 1 if worker works in Detroit; 0 otherwise (the omitted category is working in any of the thirty-one smaller cities)

WITH PARENT = 1 if the worker lives with his parents and gives them his wages; 0 otherwise

% OWN GROUP = the percent of the worker's coworkers who were born in the same nation as the worker

% PARENTS' GROUP = the percent of the worker's coworkers whose parents were born in the same nation as the worker's parents

FOREMAN = 1 if there is a foreman in the firm who is born in the same nation as the worker

FOREMAN (PARENT) = 1 if there is a foreman in the firm whose parents were born in the same nation as the worker's parents

FOREMANHIGH = 1 if the percent of the foremen born in the same nation as the worker is greater than the percent of the workers in the firm born in that nation

FOREMANHIGH (PARENT) = 1 if the percent of the foremen whose parents were born in the same nation as the worker's parents is greater than the percent of the workers in the firm whose parents were born in that nation

% OWN IN OCC = the percent of the worker's coworkers in the same occupation who were born in the same nation as the worker

% PARENT IN OCC = the percent of the worker's coworkers in the same occupation whose parents were born in the same nation as the worker's parents

UNSKILLED, SEMISKILLED, SKILLED, ARTISAN, and MANAGERIAL are a set of dummies for the worker's occupation

DAVID BUFFUM and ROBERT WHAPLES, Assistant Professor, Department of Economics, College of the Holy Cross, and Assistant Professor, Department of Economics, Wake Forest University. We wish to thank Joan Hannon, Marty Eisenberg, Robert Margo, Claudia Goldin, Hank Gemery, Richard Sutch, Price Fishback, Janice Madden, Paul Taubman, Nora Faires, Bill English, and Gerry Friedman for comments, as well as the participants of the Cliometrics Sessions at the ASSA meetings, at the Tenth North American Labor History Conference and the economics department seminars at Chicago, Lehigh, Northwestern, Ohio State, Wake Forest, and Washington (St. Louis). We also thank Roger Ransom and Richard Sutch for supplying us with additional data.