L'INSEE/GENES ADRES

Employer Size and the Wage Structure in U.S. Manufacturing Author(s): Steven J. Davis and John Haltiwanger Source: Annales d'Économie et de Statistique, No. 41/42, La microéconométrie de la gestion des ressources humaines: Etudes internationales des pratiques d'entreprises / The Microeconometrics of Human-Resource Management: Multinational Studies of Firm Practices (Jan. - Jun., 1996), pp. 323-367 Published by: L'INSEE/GENES on behalf of ADRES Stable URL: http://www.jstor.org/stable/20066474 Accessed: 14/08/2009 00:13 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=linsgen. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

L'INSEE/GENES and ADRES are collaborating with JSTOR to digitize, preserve and extend access to Annales d'Économie et de Statistique.

http://www.jstor.org

ANNALES D'?CONOMIE ET DE STATISTIQUE. - N? 41/42 ?

1996

and Size Employer the Wage Structure in U.S. Manufacturing Steven J. DAVIS, John HALTIWANGER* - We size and how ABSTRACT. study how the hourly wage structure varies with establishment Our study combines and within-plant components. breaks down into between-plant wage dispersion sector in 1982. data for the U.S. manufacturing and establishment household 1.Wage dispersion falls sharply with establishment size for nonproduction workers and mildly for production workers. inwage dispersion often mask even sharper differences in the dispersion of 2. Size-class differences and in the "skill prices" on those characteristics. wages generated by observable worker characteristics 3. In terms of dispersion m predicted log wages, tends to rise with establishment worker heterogeneity size; are more in the union sector, but only at plants with workers much homogeneous production 1,000 or more workers. 4. Unobserved factors generate sharply greater wage dispersion at smaller establishments. across establishments 5. The variance inmean wages for 59% of total variance. Within accounts for a mere 2%. variance among production workers accounts plant wage account for about one-fourth of the total between 6. Mean wage differences by size of establishment plant variance of wages. 7. Between-plant falls sharply with establishment for the size, entirely accounting wage dispersion size to overall wage dispersion. relationship of establishment negative Guided by these and other empirical findings, we assess several hypotheses about the determination of the wage structure.

Taille de l'industrie

l'employeur am?ricaine

et

structure

des

salaires

dans

le salaire Nous ?tudions comment horaire varie en fonction de la taille des R?SUM?. en une part inter-?tablissements et comment la dispersion des salaires se d?composent ?tablissements une et et d'entreprises Notre ?tude montre qu'? partir de donn?es de m?nages part intra-?tablissement. sur l'industrie am?ricaine en 1982. 1. La dispersion des salaires diminuent fortement avec la taille de l'?tablissement pour les employ?s et l?g?rement pour les ouvriers. 2. Ces diff?rences de dispersion masquent des ?carts de dispersion g?n?r?s par les caract?ristiques et le prix de la qualit? du travail. individuelles observables 3. En terme de dispersion les salaires pour esp?r?s. avec la taille de l'?tablissement. des salari?s tend ? augmenter L'h?t?rog?n?it? Les ouvriers sont beaucoup dans le secteur syndiqu?, mais seulement pour les plus homog?nes ?tablissements de 1000 salari?s ou plus. 4. Les caract?ristiques non sont individuelles observables d'une plus grande part de responsables au sein des ?tablissements la dispersion les plus petits. 5. La variance des salaires moyens entre ?tablissements 59 % de la variance totale. pour compte A l'oppos?, la variance du salaire des ouvriers au sein de l'?tablissement ne rend compte que de 2 % de cette dispersion totale. 6. La taille de l'?tablissement rend compte d'un quart de la variance des salaires moyens entre ?tablissements. 7. La dispersion de salaire entre ?tablissement chute fortement avec la taille de l'?tablissement, rendant compte en totalit? de la relation n?gative entre la taille des ?tablissements et la dispersion totale. Guid? par ces r?sultats ainsi que par d'autres nous ?valuons r?sultats empiriques, plusieurs sur la d?termination de la structure de salaire. hypoth?ses

* S. J. DAVIS: University of Chicago; J. HALTIWANGER: University of Maryland. Both authors are also research associates at the Center for Economic Bureau of the Census. In Studies, the LRD data for this study, we have greatly benefited from the assistance of Bob preparing Robert McGuckin and other Census Bureau employees Bechtold, Tim Dunne, James Monahan, at the Center for Economic We also thank Kevin Murphy Studies. on for helpful discussions CPS data. Laura Power and Catherine research excellent assistance. We Buffington provided research support provided a Joint Foundation, gratefully acknowledge by the National Science Statistical between the Census Bureau and the University of Maryland, and the Agreement Graduate School of Business at the University of Chicago.

1 Introduction

This strong empirical pay substantially Larger employers higher wages. regularity holds up across studies that span several decades, many countries and regions, and a broad range of industries and occupations for l. Evidence sector points to sharply rising size-wage differentials the U.S. manufacturing over

time.

on the subject A few facts from our previous work and (Davis Haltiwanger of size-wage differentials, their [1991]) convey the magnitude to the general increase in U.S. wage growth over time, and their relationship that has occurred since the 1970s. Among U.S. manufacturing inequality the wage differential between plants with more than production workers, rose from $3.40 an and plants with 20 to 49 employees 5,000 employees hour (1982 dollars) in 1967 to $6.31 an hour in 1986. These size-wage and the increase over time, are huge relative to the overall differentials, mean production worker wage of $8.18 an hour in 1967 and $9.18 an hour in 1986. Furthermore, the rise in the size-wage accounts for a premium fraction of the in overall Between 1976 wage inequality. significant growth account for 40% of the increase in and 1985, rising size-wage differentials wage dispersion among U.S. manufacturing between-plant plants, and the turn accounts in the 48% overall for of component growth in between-plant workers. wage variance among manufacturing a in this paper, the dispersion of wages also exhibits to In size. overall wage employer particular, pronounced relationship and size for both production falls sharply with establishment dispersion in sector. workers the U.S. This manufacturing nonproduction empirical of wage component entirely reflects a smaller between-plant relationship wage among larger establishments. Within-plant dispersion dispersion size. actually rises with establishment As

we

show

the empirical The factors underlying involving regularities employer are not well A careful understood. size and wages study by Brown several hypotheses about the source of mean and Medoff [1989] explores of different sizes: (i) larger employers among employers wage differentials offer inferior working hire higher quality workers; (ii) larger employers or pay are more conditions; (iii) larger employers likely to be unionized to have avoid unionization; (iv) larger employers greater rents higher wages in to share with workers; (v) larger employers pay higher efficiency wages to have and (vi) larger employers response relatively monitoring problems; to receive empirical low applicant-to-vacancy ratios. The only hypothesis size classes, among quality differences support in their study is worker account for roughly one-half of observed differentials. which size-wage

1. Early [1989] Chapter

324

and Medoff Brown include Moore [1954] and Lester [1967]. [1911], Rehn on the size-wage et al. [1990] review much of the evidence relationship. also offer more attractive fringe benefits. 5 in Brown et al. shows that larger employers

studies

and Brown

s findings Thus, Brown and Medoff mean wage differentials by employer

indicate that nearly half of observed size remains unexplained.

of manufacturing how the distribution In this paper, we investigate size classes. We use establishment differs workers' among hourly wages on information characteristics, hourly earnings, demographic person-level status and employer size to quantify how (i) differences union coverage in the wages in observable worker and job characteristics, (ii) differences and (iii) differences in unobserved characteristics associated with observable factors contribute to differences among employer size classes in the structure data data with establishment-level of wages. We also combine person-level to decompose into between-plant and within-plant total wage dispersion size class, an We carry out this decomposition components. by employer and within-plant exercise that enables us to infer how between-plant wage dispersion

varies

by

size of establishment.

of the connection between that a better understanding anticipate structure of the distribution will shed additional size and the wage employer size and mean wages, between employer the focus light on the connection on the In of most research addition, size-wage previous relationship. involve the connection several interesting and largely unexplored hypotheses size and the dispersion of wages between both among workers, employer between plants and within plants: (i) unions compress wages within plants, or to mimic union wage and larger plants are more likely to be unionized as a union avoidance tactic; (ii) large employers policies rely more heavily on ability-sensitive of pay for managerial methods (iii) large employees, are more likely to use a standardized technology employers requiring workers; (iv) large employers carry out a greater variety homogeneous of tasks and accordingly have a greater diversity of workers; (v) larger more on rate standard pay rely employers heavily policies, leading to greater on observable worker characteristics); of wages (conditional compression and (vi) the heterogeneity in average workforce quality is greater among small employers than among large employers. We provide direct and indirect evidence related to these and other hypotheses about the connection between structure. size and the wage employer We

several sources of data. We obtain person-level study combines and other to the May data the Employee Benefits Supplement from wages to 1983 Current Population which is matched the March 1983 Survey (CPS), Annual Demographic File. The Benefits Supplement includes information on size and union status, and the Annual Demographic File includes employer information on hourly earnings in 1982. We draw upon the 1982 Census of to obtain data on individual manufacturing Manufactures establishments. Our

structure hypotheses the wage to relate most directly we our attention to the manufacturing restrict sector. within-plant dispersion, For this sector, we can combine CPS data with Census of Manufactures data in the Longitudinal Research Database total (LRD) to decompose into and wage dispersion components, between-plant within-plant using the in Davis and Haltiwanger Our [1991]. methodology developed previous work on this decomposition devoted attention and effort to considerable for possible biases induced by the exclusion of administrative adjustments establishments from the LRD sampling frame. In this study, we augment Since

several

of

EMPLOYER SIZEAND THEWAGE STRUCTURE

325

to incorporate the LRD administrative and auxiliary establishments in a data set with observations the manufacturing sector, thereby yielding on the universe of U.S. manufacturing in 1982. establishments The are drawn from the data on administrative and auxiliary establishments Census Bureau's Standard Statistical Establishment List (SSEL). Henceforth, to the LRD to reflect an references in this paper should be understood LRD version of the that includes and administrative augmented auxiliary establishments. We note that several other recent studies stress the role of establishment size in establishment-level and productivity behavior. wage, employment and Samuelson Dunne, Roberts [1989] show that plant failure rates are in plant size. Davis and Haltiwanger [1992] and Davis, sharply decreasing Haltiwanger and Schuh that the dispersion of [1996] present evidence establishment-level growth rates is inversely related to plant size. Baily, show that plant-level is an Hulten and Campbell [1992] productivity function of size. Dunne and Schmitz that find [1991] plant increasing are more to advanced and that this likely adopt technologies, larger plants accounts behavior for of the size in technology difference part adoption Troske [1994] shows that capital intensity differences wage differential. for mean wage differences among play an important role in accounting these and many other studies size employers. Taken different together, size is a key correlate of wage, employment and make clear that employer behavior. productivity

Basic Patterns

in the Size-Wage

Relationship

in 1982 by establishment 1 and 2 depict mean hourly wages Figures in the manufacturing workers and nonproduction size class for production 1% of each representing sector. The figures consider 100 size classes, construct the these from We augmented figures employment. manufacturing on all U.S. manufacturing establishments LRD, which contains observations in 1982. Thus, the data set contains approximately 300,000 manufacturing about 18.5 million workers. This large number of establishments employing enables us to draw a much sharper picture observations establishment-level than previous research. size and mean wages of the relationship between result from the sharper picture, and two points, Some useful refinements are large differentials in particular, emerge clearly. First, mean size-wage For production workers, workers. and nonproduction for both production the mean wage differential between plants in the 10th and 90th size classes is $5.05 per hour, which equals about 62% of the overall mean production is sharply nonlinear the size-wage worker wage. (in Second, relationship and nonproduction and it differs between production size class percentiles), with rise systematically For production workers. size, workers, wages although the slope is much steeper in the upper half of the size distribution. For nonproduction workers, wages rise steeply with size in the first 10 size

326

1

Figure

HOURLYWAGES BY PLANTSIZE CLASS PRODUCTION WORKERS,

40

Figure

1982

60 50 SIZE CLASS

2

HOURLYWAGES BY PLANT SIZE CLASS NONPRODUCTION WORKERS,

1982

18

16

14 LU H o i

12 H

10 4

I I I I I I 11I I I I I 1 I 11I I IM I I I I I I I I I I I IM I 1I I I I 11I I I I I I I 11I I I I I I I 11IM I IN 11l IM I I I IM IM I I I I 11I I I I I I I II I 20 30 50 60 70 80 90 10 40 100

SIZE CLASS

EMPLOYER SIZEAND THEWAGE STRUCTURE

'

327

fall slightly between the 15th and 50th size class, and then rise the 50th and 100th size classes 2. sharply between of variation of hourly 3 and 4 depict the mean and coefficient Figures and size class for establishment wages manufacturing production by We construct from the these 1982 workers. earnings figures nonproduction matched CPS files. These two figures display a data in the 1983 March-May the wage mean and wage dispersion relationships striking contrast between to establishment rise sharply with size, wage size. Although mean wages from workers. Moving dispersion falls sharply, especially for nonproduction to the coefficient 1-24 with 1000+ with employees, employees plants plants of variation for hourly wages falls from 44% to 38% for production workers workers. and from 66% to 49% for nonproduction classes,

Figure

3

MEANWAGES AND WAGE DISPERSION BY PLANTSIZE (PRODWORKERS) 11 ..-_-__

10.5-

w.t*f

^^^^

\ \

10-1 z 9-5- \ /

0.43

/