Thursday, June 30, 2022
Positivism versus antipositivism
Marx criticized Kant.
I'd say the essential Marxist critique of positivism is that the latter is trapped in an INDIVIDUALIST model of the scientific observer . THE empiricist is an individual . Newton's standing on the shoulderS of giants is ignored. All science is socially determined.
Sent from my iPhone
On Apr 8, 2016, at 8:50 PM, Jim Farmelant wrote:
One curious aspect of the postivist/antipositivist debates within the social sciences is that they cut across a lot of other ideological differences within the social sciences. Among Marxists, such debates have been going for over a hundred years. Bukharin, for instance wrote a treatise on historical materialism that was explicitly positivist in his approach. Antonio Gramsci famously wrote a review that panned that book, in which Gramsci took a strongly antipositivist stance, as did Lukacs. Likewise, among non-Marxists, similar debates have been going on for a long time. People like Comte, Spencer, and Durkheim, all defended positivist approaches to social science, While people like Dilthey, and Weber were more or less critical of positivism in the social sciences, as have people like Scheler, Levi-Strauss, and Geertz.
You can see these positivist/antipositivist debates going on in different social science disciplines and within different ideological camps. Thus, within economics, Milton Friedman was a notable defender of a positivist approach to economics, whereas, some of the members of the Austrian School, like Ludvig von Mises and Friedrich Hayek were notable antipositivists. Yet, Friedman, Mises, and Hayek, all shared similar political outlooks. At the same time, the British socialist economist Joan Robinson, a famoust Left Keynesian, in her 1962 book, Economic Philosophy, used positivism to ground her critique of mainstream economics.
Both von Mises and Hayek, were originally reacting against the writings in defense of socialist economic planning that were penned by the socialist economist Otto Neurath, who was also one of the founders of the Vienna Circles of logical positivists. For Hayek, in particular, the critique of positivism in economics and the social sciences generally was a crucial part of the struggle against socialism.
Meanwhile, the Frankfurt School, which was Marxist, was also very notably antipositiivst. For them, positivism was one of the fundamental props of contemporary capitalist ideologies so for them, the critique of positivism was necessary for the struggle against capitalism.
https://www.marxists.org/archive/bukharin/works/1921/histmat/
https://www.marxists.org/reference/subject/philosophy/works/us/friedman.htm
http://people.stfx.ca/jcook/2010-11/The%20Scientific%20Conception%20of%20the%20World.pdf
https://archive.org/details/counterrevolutio030197mbp
http://marxistupdate.blogspot.com/2012/01/lukacs-on-bukharins-theory-of.html
https://archive.org/details/EconomicPhilosophy
Jim Farmelant
http://independent.academia.edu/JimFarmelant
http://www.foxymath.com
Learn or Review Basic Math
Engels anticipated No Aether discovery in Physics
chronological thread
From: Charles Brown
Subject: Re: [marxism-thaxis] Engels anticipates No Aether in Physics
Date: Wed, 20 Jul 2016 22:59:55 -0400
Sent from my iPhone
On Apr 30, 2015, at 10:59 AM, Charles Brown wrote:
Consider Engels principle that
" There is nothing but matter, and its mode of existence is motion " in relation to Newton's First Law of motion
"When viewed in an inertial reference frame, an object either remains at rest or continues to move at a constant velocity, unless acted upon by an external force.[2][3] http://en.m.wikipedia.org/wiki/Newton%27s_laws_of_motion;
And subsequent developments in physics in the Michaelson-Morley experiment disproving the existence of "the ether " or Absolute Rest;and
Einstein's Theory of Relativity in which there is no Absolute Rest but only Relative forms if motion and a sort of Absolute Motion in the form of Light .
Engels formulation is that there is no matter at rest. Matter's only mode of existence is motion, contra the great physicist Newton, who asserts that some objects are at rest in his Law. And physics had the concept of the aether as absolute rest until it was disproven. No absolute rest is basic to Einstein's famous revolutionary theory.
So, Engels anticipates the critique of Newton in physics with his "absolute " motion dialectical principle . How's that for materialism as science, non-mystical dialectics ?
Sent from my iPhone
Re: [marxism-thaxis] Engels anticipates No Aether in Physics, Charles Brown, 07/21/2016
Only 4 of 55 African leaders attend Zelensky call, showing neutrality on Ukraine and Russia.
6/22/2022
Only 4 of 55 African leaders attend Zelensky call, showing neutrality on Ukraine and Russia. By: Benjamin Norton
0 COMMENTS
France and Germany pressured African Union leaders for months to join a brief Zoom call with Ukraine’s Volodymyr Zelensky. 51 of 55 African heads of state (93%) boycotted the meeting, showing clear neutrality over the Western proxy war with Russia.
Picture Only four of 55 African leaders join a Zoom call with Ukraine's Zelensky on June 20, 2022 Western governments have tried to rally the nations of Africa to join their war on Russia. But the vast majority of the continent has ignored their pressure campaign.
For months, Ukraine attempted to organize a video conference between the African Union and Western-backed leader Volodymyr Zelensky.
France and Germany put heavy pressure on African governments to attend the Zoom call, which was held on June 20.
The conference ended up being a total failure, however. The heads of state of just four of the 55 members of the African Union joined the meeting.
In other words, 93% of the leaders of the African continent did not attend the video conference with Zelensky.
This was a clear sign of Africa’s overwhelming neutrality in the proxy war between the West and Russia.
France’s major newspaper Le Monde described Zelensky’s video call as “an address that the African Union (AU) has delayed for as long as possible and has been keen to keep discreet, almost secret.”
Ukraine had tried to organize the conference since April, but the AU had repeatedly pushed it back.
Le Monde noted that “the organization of the simple video message illustrates the tense relationships between Mr. Zelensky and the leaders of the continent,” who are “sticking to a neutral position.”
Citing an internal source, The Africa Report identified the very few African heads of state who attended the call as Senegal’s President Macky Sall, Côte d’Ivoire’s President Alassane Ouattara, and the Republic of the Congo’s President Denis Sassou Nguesso.
Also at the video conference was Mohamed al-Menfi, the leader of the Libyan Presidential Council, which is recognized by some countries as a legitimate government, although this is disputed by many nations, and Libya has remained territorially divided since a 2011 NATO war destroyed the central state.
At the meeting with Zelensky, these three or four heads of state were joined by Moussa Faki, a politician from Chad who serves as chair of the African Union, and some lower level diplomats of other countries.
The African Union apparently tried to keep the conference as quiet as possible. It did not post anything about the call on its official website. It did not tweet about the meeting either.
The only official recognition of the call came from Faki, in a lone tweet, in which he cautiously “reiterated the AU position of the urgent need for dialogue to end the conflict to allow peace to return to the Region and to restore global stability.”
The United States and European Union frequently claim that they are acting on behalf of the “international community,” but events like this demonstrate that when Washington and Brussels say international community, they actually just mean the roughly 15% of the global population in the West and their loyal allies in Australia, New Zealand, South Korea, and Japan.
Multipolarista detailed in a report in March how the vast majority of the world’s population, which resides in the Global South, has remained neutral over the Western proxy war in Ukraine.
Countries with some of the largest populations on Earth, such as China, India, Pakistan, Brazil, Ethiopia, Bangladesh, Mexico, and Vietnam, have remained neutral.
Many more nations in the Global South, such as South Africa, Iran, Venezuela, Cuba, Nicaragua, North Korea, and Eritrea, have openly blamed NATO and the United States for causing the war in Ukraine.
Establishment British newspaper The Guardian, which is closely linked to UK intelligence agencies, published an article in March reluctantly acknowledging that many African countries “remember Moscow’s support for liberation from colonial rule, and a strong anti-imperialist feeling remains.”
The report noted that a significant number of African leaders are “calling for peace but blaming Nato’s eastward expansion for the war, complaining of western ‘double standards’ and resisting all calls to criticise Russia.”
It conceded that nations like South Africa, Zimbabwe, Angola, and Mozambique, “are still ruled by parties that were supported by Moscow during their struggles for liberation from colonial or white supremacist rule.”
Russia today also has important trade relations with Africa. As one of the world’s top producers of wheat, Russia is a significant source of food for the continent.
While food insecurity is an endemic problem in formerly colonized nations in Africa that were ravaged by centuries of Western imperialism, the United States has threatened to make this crisis even worse.
The New York Times reported that the US government is pressuring food-insecure countries in Africa not to buy Russian wheat. Author Benjamin Norton is a journalist, writer, and filmmaker. He is the founder and editor of Multipolarista, and is based in Latin America. // Benjamín Norton es un periodista, escritor, y cineasta. Es fundador y editor de Multipolarista, y vive en Latinoamérica. This article was republished from Multipolarista.
France and Germany pressured African Union leaders for months to join a brief Zoom call with Ukraine’s Volodymyr Zelensky. 51 of 55 African heads of state (93%) boycotted the meeting, showing clear neutrality over the Western proxy war with Russia.
Picture Only four of 55 African leaders join a Zoom call with Ukraine's Zelensky on June 20, 2022 Western governments have tried to rally the nations of Africa to join their war on Russia. But the vast majority of the continent has ignored their pressure campaign.
For months, Ukraine attempted to organize a video conference between the African Union and Western-backed leader Volodymyr Zelensky.
France and Germany put heavy pressure on African governments to attend the Zoom call, which was held on June 20.
The conference ended up being a total failure, however. The heads of state of just four of the 55 members of the African Union joined the meeting.
In other words, 93% of the leaders of the African continent did not attend the video conference with Zelensky.
This was a clear sign of Africa’s overwhelming neutrality in the proxy war between the West and Russia.
France’s major newspaper Le Monde described Zelensky’s video call as “an address that the African Union (AU) has delayed for as long as possible and has been keen to keep discreet, almost secret.”
Ukraine had tried to organize the conference since April, but the AU had repeatedly pushed it back.
Le Monde noted that “the organization of the simple video message illustrates the tense relationships between Mr. Zelensky and the leaders of the continent,” who are “sticking to a neutral position.”
Citing an internal source, The Africa Report identified the very few African heads of state who attended the call as Senegal’s President Macky Sall, Côte d’Ivoire’s President Alassane Ouattara, and the Republic of the Congo’s President Denis Sassou Nguesso.
Also at the video conference was Mohamed al-Menfi, the leader of the Libyan Presidential Council, which is recognized by some countries as a legitimate government, although this is disputed by many nations, and Libya has remained territorially divided since a 2011 NATO war destroyed the central state.
At the meeting with Zelensky, these three or four heads of state were joined by Moussa Faki, a politician from Chad who serves as chair of the African Union, and some lower level diplomats of other countries.
The African Union apparently tried to keep the conference as quiet as possible. It did not post anything about the call on its official website. It did not tweet about the meeting either.
The only official recognition of the call came from Faki, in a lone tweet, in which he cautiously “reiterated the AU position of the urgent need for dialogue to end the conflict to allow peace to return to the Region and to restore global stability.”
The United States and European Union frequently claim that they are acting on behalf of the “international community,” but events like this demonstrate that when Washington and Brussels say international community, they actually just mean the roughly 15% of the global population in the West and their loyal allies in Australia, New Zealand, South Korea, and Japan.
Multipolarista detailed in a report in March how the vast majority of the world’s population, which resides in the Global South, has remained neutral over the Western proxy war in Ukraine.
Countries with some of the largest populations on Earth, such as China, India, Pakistan, Brazil, Ethiopia, Bangladesh, Mexico, and Vietnam, have remained neutral.
Many more nations in the Global South, such as South Africa, Iran, Venezuela, Cuba, Nicaragua, North Korea, and Eritrea, have openly blamed NATO and the United States for causing the war in Ukraine.
Establishment British newspaper The Guardian, which is closely linked to UK intelligence agencies, published an article in March reluctantly acknowledging that many African countries “remember Moscow’s support for liberation from colonial rule, and a strong anti-imperialist feeling remains.”
The report noted that a significant number of African leaders are “calling for peace but blaming Nato’s eastward expansion for the war, complaining of western ‘double standards’ and resisting all calls to criticise Russia.”
It conceded that nations like South Africa, Zimbabwe, Angola, and Mozambique, “are still ruled by parties that were supported by Moscow during their struggles for liberation from colonial or white supremacist rule.”
Russia today also has important trade relations with Africa. As one of the world’s top producers of wheat, Russia is a significant source of food for the continent.
While food insecurity is an endemic problem in formerly colonized nations in Africa that were ravaged by centuries of Western imperialism, the United States has threatened to make this crisis even worse.
The New York Times reported that the US government is pressuring food-insecure countries in Africa not to buy Russian wheat. Author Benjamin Norton is a journalist, writer, and filmmaker. He is the founder and editor of Multipolarista, and is based in Latin America. // Benjamín Norton es un periodista, escritor, y cineasta. Es fundador y editor de Multipolarista, y vive en Latinoamérica. This article was republished from Multipolarista.
Social and Cultural Anthropology: A Very Short Introduction. Review article by: Thomas Riggins
https://www.midwesternmarx.com/articles/social-and-cultural-anthropology-a-very-short-introduction-review-article-by-thomas-riggins
Picture This short book (155 pages) by John Monaghan and Peter Just from Oxford University Press is a really good introduction to this subject. Although it does not transcend a bourgeois worldview it will give you the needed information to cope with literature in this field.
The book cuts to the heart of a scientific discipline too many people shy away from as too difficult, remote, or bazaar. The book is part of an Oxford University series designed to make any subject matter easily accessible to anyone interested enough to read a short, very short, introductory text that will, nevertheless, give the reader a grasp of the main points of the subject under discussion along with excellent bibliographic references for further independent study. This volume will be a welcome addition to anyone's library on ‘Third World’ issues, or more importantly, on those of the ‘Fourth World’ of indigenous peoples restricted to the margins of the world imperialist system.
Anthropology deals, historically least, with the cultures and social institutions of pre-industrial, pre-state, and pre-literate peoples. This book is designed to provide an understanding as to how such societies function, and how they relate to, and are being destroyed by, the modern world system of imperialist globalization (my term, not the authors).
I suggest this book for all activists with no prior exposure to anthropology: the more we know about Third and Fourth World peoples, especially the indigenous peoples of Africa, Asia, the Americas and the Pacific islands, the more we can understand the scope of the struggle against imperialism and the way to win allies against a common enemy.
Social and Cultural Anthropology, useful as it is as an introduction to a complicated social science, is yet a product of the bourgeois US university system. I will outline some of the problems that Marxists must keep in mind while reading this book. First, you might not be getting the names of real persons and places in anthropology books. This is because anthropology is used by the US and other repressive governments to keep tabs on groups that may be a “problem.” Wounded Knees are still a daily occurrence for indigenous peoples, and the same US government which sponsored the original is tacitly behind the replicas throughout the world.
Second, the author’s discussion of “Cultural Relativism” is muddled and weak. They seem to confuse two different concepts: Different cultures have different values, and all cultural values are equal.
For example, the authors discuss the practice of female genital mutilation (they call it “female circumcision”) practiced by the Hofriyati people of the Northern Sudan. They say: ” We may find the consequences of such practices repellent, but we are hard pressed to find a moral basis for advocating its suppression that does not also violate the cultural autonomy of the Hofriyati. One wonders , ultimately, if it is logically possible to simultaneously subscribe to both the notion of universal human rights and a belief in the relativity of cultures.”
I remember being told in a carpet factory outlet in India that objection to children laboring sixteen hours a day in a factory was Western cultural inference with Indian traditions. I wonder when maximizing surplus value in factories became part of the Indian tradition. In any event, I refer the authors to another book in the Oxford series, Logic: A Very Short Introduction, for the resolution of their logical conundrum.
Third, the authors use a lot of “post-modern” terms, which they don’t define and are quite meaningless, such as the “post-industrial” West. Is the West also “post-pollution”?
There is an interesting section on Marxism and ‘’neoevolutionary anthropology’’ where the older categories of Morgan and Engels — savagery, barbarism, civilization — are replaced by the now more generally accepted “four basic patterns of human society”— namely, foraging societies, tribal societies, chiefdoms, and states.
This scheme is also the one used in a book previously reviewed here, Guns, Germs, and Steel, and both should be compared to The Origin of the Family, Private Property, and the State by Engels which was based on Henry Lewis Morgan’s Ancient Society.
Other subjects discussed in this book are religion as it relates to social structure, gender as defined in different cultures, and the “positive” side of relativism.
“When someone begins a peroration with the phrase ‘but of course it’s human nature to…’ start looking for the exit! Because what you are about to hear will most likely reflect the speaker’s most deeply held prejudices rather than the product of a genuine cross-cultural understanding. Every time anthropologists have attempted to generate universal rules governing human behavior, the rules have either been proven empirically wrong or are so trivial as to be uninteresting.”
I suggest this short book to anyone who has an interest in other cultures and peoples outside the ambit of “Western Civilization” — of course with the caveat any Marxist must keep in mind when reading a bourgeois, even a progressive bourgeois, work, namely to be en garde.
Wednesday, June 29, 2022
The Employer Size-Wage Effect
Charles Brown ( NO RELATION -CB); James Medoff
The Journal of Political Economy, Vol. 97, No. 5. (Oct., 1989), pp. 1027-1059. Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28198910%2997%3A5%3C1027%3ATESE%3E2.0.CO%3B2-E
The Journal of Political Economy is currently published by The University of Chicago Press.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. 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/journals/ucpress.html.
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.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.
http://www.jstor.org Mon Jan 14 17:30:07 2008
The Employer Size-Wage Effect
Charles Brown
Unzuer\2tv of Mlchtgnn and .Vattonal Hurmtc of Econo~nlcKrsrnrch
James Medoff
Hnnmrd CInzuer\2t) crnd Satzonal Bureau (4Econom2c Retecrrch
We consider six explanations for the positive relationship between employer size and wages: large employers (1) hire higher-quality workers, (2) offer inferior working conditions, (3) make more use of high wages to forestall unionization, (4)have more ability to pay high wages, (5)face smaller pools of applicants relative to vacancies, and (6)are less able to monitor their workers. We find some support for the first of these, but there remains a significant wage premium for those working for large employers.
There is much evidence that "large" employers pay more than "small" employers even when their union status is the same (Lester 1967; Masters 1969; Antos 1981; Mellow 1982; Atrostic 1983; Oi 1983). There is, however, much less information that can help us answer a number of key questions concerning this wage differential: Is it com- pany size or establishment size that matters for wages, or does each have an independent effect? If employer size is treated in a continu- ous fashion, precisely how big is the size-wage effect? How much of
U'e have benefited from helpful comments by seminar participants at SBER and at Princeton, Chicago, and North Carolina State ~rniversitiesand by John Caren, i\ndrew Weiss, and our referees. Sicole Garris, Cheryl Hansen, Sandy Korenman, Alan Krue- ger, Nancy Lemrow, Michael hlandel, Marsha Silverberg, and Martin Van Denburgh provided invaluable assistance. Support from the National Science Foundation (grant 2342) and the (:omputer Science Centers at the University of Maryland (where Brown worked on earlier drafts of the paper) and the University of Michigan are gratefully acknowledged.
1028 JOURNAL OF POLITICAL ECONOMY
this employer size-wage differential can be explained by the fact that employers who are larger hire better workers? How much can be explained by a size differential in working conditions? How much by the fact that, in the nonunion sector, larger employers do more to avoid unions than smaller employers? How much by a differential in product market power? Why else might larger employers pay more?
This study addresses each of these questions. Section I discusses the four factors that dominate previous research on determinants of the employer size-wage differential: labor quality, working conditions, union avoidance, and product market power.
The size-wage differential is one of the key differentials observed in labor markets. It is particularly interesting because, unlike the union wage differential, it exists in the absence of an obvious agent, one of whose goals is its existence. Hence, if employers of different sizes pay very differently for the same quality of labor working in a similar environment, there is no readily available deus ex machina to save the day: our knowledge of the labor exchange must ultimately be relied on.
Section I1 provides the stylized facts of the matter. The evidence reveals that the size-wage effect is quite large: the wage gain associ- ated with moving from an employer whose size is one standard devia- tion below average to an employer one standard deviation above aver- age is roughly the same as the gain associated with moving from a nonunionized to a unionized employer. Company size and establish- ment size have independent effects on pay. Finally, the findings pre- sented strongly suggest that, while a size differential in labor quality can explain about one-half of the total size-wage differential, the other three factors under consideration can explain little of the re- mainder. Since the residual size-wage effect is large, the question of why size matters for wages is much more perplexing than it may have appeared at first blush.
Section I11 presents a number of additional hypotheses about the origins of employer size-pay differentials, and the likely explanatory power of each of these theories is assessed. Section IV presents con- clusions.
I. Some Traditional Explanations
Both neoclassical and institutional labor economists have offered ex- planations of why larger employers pay more than smaller employers. As would be expected, the neoclassicists have focused on size differ- entials in labor quality or working conditions. While the institutional- ist approaches are more varied, they often turn to factors such as
S I Z E - W A GE FEF E C T 1 0 2 9
union avoidance and product market power. While each of the expla- nations alluded to is plausible, neither their individual nor their col- lective power has yet been tested.
The Neoclassical Exfllanations
The theory of "compensating differentials" or "equalizing differ- ences" is at the heart of neoclassical labor economics. It is for this reason that many of the existing discussions of size-wage differentials focus on size differentials in the quality of labor or the conditions of work.
T h e labor quality explanation of the size-wage relationship can be simply put: larger firms or establishments employ higher-quality workers. There are several reasons why larger employers might make greater use of higher-quality labor, all else the same. Greater capital intensity of larger establishments and capital-skill complementarity provide one explanation (Hamermesh 1980, p. 386). An alternative is presented by Oi (1983),who suggests that large firms employ higher- quality workers in order to reduce the costs of monitoring a given quantity of labor services. In deriving this result, Oi makes the very strong assumption that greater entrepreneurial ability (which is what generates larger firms in the model) increases the quantity of decision making that can be achieved in an hour of the entrepreneur's time, but it does not affect the number of workers whose output can be monitored per hour by the entrepreneur.'
The labor quality explanation of the size-wageeffect lends itself to a number of statistical tests. The first involves the estimation of wage equations with cross-sectional data on individuals. Very simply put, the analysis asks whether or not the estimated size-wage effect can be explained in terms of measured dimensions of labor quality. While unmeasured dimensions of quality clearly exist, one might hope that they will be correlated with measured variables such as schooling, age, and the like.
The labor quality explanation can be addressed in a different fash- ion with longitudinal data, by comparing the wages of the same indi- vidual when he or she is working for different-sized employers. To the extent that unmeasured dimensions of worker quality are fixed over time, looking at wage rate change as a function of change in employer size (and other measured factors) will give an estimate of the size-wage effect that is not biased by constant dimensions of labor
For a model in which firm size is related to the ability of the entrepreneur but not necessaril) to the abilit\ of other workers, see Rosen (1982).
1030 JOURNAL OF POLITICAL ECONOMY
quality. It should be noted, however, that the downward bias resulting from classical measurement error in the size variable will be exacer- bated by differencing if the ratio of error variance to true variance is greater for the change in size than for its level.
Other pieces of evidence may help to distinguish among competing labor quality hypotheses. For instance, the explanation that links skill to scale on the basis of capital-skill complementarity tends to take establishment size as the relevant measure of scale and hence predict a relationship between wages and establishment size. Oi's monitoring model focuses on the effects of firm size. While the correlation be- tween firm and establishment size is strong enough to produce similar results when only one measure is used, including both firm and estab- lishment size can clearly refine the "stylized facts" that must be ex- plained.
If the higher wages of larger employers are due to differences in worker quality, then those working for larger employers would earn no more than they would earn elsewhere. This would imply that the quit rates of larger enlployers would be no different from the quit rates of smaller employers when wages and nonwage benefits (includ- ing a larger menu of potential jobs) are held fixed. Hence, informa- tion on size differentials in quit rates, company tenure, and "job" tenure can shed light on the size-wage effect puzzle.
While the details differ, each of the labor quality hypotheses pre- sented above has larger employers and smaller employers paying the same wage for workers of given quality. Alternatively, undesirable working conditions generally associated with larger workplaces- such as greater reliance on rules and less freedom of action and scheduling (Masters 1969; Stafford 1980), more impersonal work at- mosphere (Lester 1967), or longer commuting (Scherer 1976, p.
111)-may force larger employers to pay higher wages to get a given quality of labor.
T h e first step in testing this compensating differentials explanation involves isolating the unattractive aspects of larger workplaces. In- cluding variables for these 'tjob characteristics" in a wage equation should reduce or eliminate the wage premium associated with em- ployer size. Unfortunately, some job characteristics are hard to mea- sure directly. However, since it seems likely that a substantial fraction of the total variation in such working conditions occurs across indus- tries and occupations, an analysis that fits wage equations without and with detailed industry and occupation controls would provide valu- able information about the validity of the working conditions expla- nation. Finally, if the compensating differentials view is correct, size should be positively related to quits if the wage is held constant but working conditions are not.
SIZE-WAGE EFFECT 1031
The Institutionnl Explnnntions
One institutionally oriented explanation for the differences in labor market behavior between large and srnall employers is that large nonunion employers act in many ways as if they were unionized in order to avoid unionization. There is considerable evidence that em- ployers that follow a strategy of "positive labor relations" to avoid unionism will pay higher wages, offer more benefits, and provide better working conditions than otherwise similar nonunion employers (see Curtin 1970, p. 60; Foulkes 1980; Freeman and Medoff 1983, p. 153).Since it is primarily the large employers that adopt such person- nel policies, one result is that union wage and benefit differentials vary inversely with size. Freeman and Medoff describe just such a variation, with a union wage differential of 5 percent for workers in firms with more than 1,000 workers compared with 22 percent for workers in firms with fewer than 100 workers. They find a similar
pattern for fringe benefits.
T h e importance of union avoidance efforts can be assessed by more
detailed investigation of the size-wage relationship. We can determine whether size-wage differentials exist even within the union sector and in occupations or industries for which there is a near-zero threat of unionism. For both already-organized and unorganizable workers, union avoidance by large employers is unlikely.
Each of the theories discussed so far assumes (or is consistent with) cost-minimizing behavior by firms. Alternatively, large firms or estab- lishments are sometimes said to engage in different labor market behavior than smaller ones because they possess product market power. One common argument is that firms with "monopoly power" may share with their workers some of the "excess" profits or rents that such power yields (Weiss 1966; Mellow 1982). However, even if large employers did use their excess profits to overpay their workers, one must still explain why they pay more than market wages and why competition for these choice jobs does not lead to a work force that is overqualified but not overpaid.
It is not clear whether the product market power explanation refers to large firms or simply to industries in which the typical firm is large. The latter view can be tested by controlling for detailed industry in wage regressions. In any case, the premise of the argument-the product demand curves of large employers are less elastic-can be checked directly.
11. Stylized Facts
The tables presented in this section shed light on the likely validity of the explanations outlined above. In addition to documenting the in-
1032 JOURNAL.OF POLITICAL ECONOMY
ability of this set of explanations to resolve the employer size-wage effect puzzle, the section provides a new set of facts with which the ultimate explanation must reckon.
Ezlzdence on the Labor Qualztj Explanation
Tables 1 and 2 relate to the "labor quality" explanation of the size- wage effect. Table 1 offers estimates of size-wage differentials based on five data files: The Current Population Survey (CPS)and Quality of Employment Survey (QES) give data for individuals, while the Survey of Employer Expenditures for Employee Compensation (EEEC), the Wage Distribution Survey (WDS), and the Minimum Wage Employer Survey (MWES) contain data for establi~hments.~
Three of these files have information on company size in addition to establishment (or location) size. These make it possible to deter- mine the wage differential associated with establishment size, with company size held constant, and vice versa. The WDS company size variable is, however, only an estimate, calculated as employment in the company's surveyed establishment(s) times the ratio of enterprise (company) to surveyed establishment sales.3~stablishmentsize is re- ported by size category (e.g.,500- 1,000 workers) in the CPS and QES (and for company size for multiestablishment companies in the EEEC); each set of categories was converted to a continuous variable using the estimated mean employment by size category and broad industry (based on Countj Business Patterns data [U.S. Bureau of the
'Our CPS sample consists of respondents to the May 1979 supplement to the Cur- rent Population Survey. T o maintain comparability with other data files, we limited our sample to private wage and salary workers. The QES, conducted by the Survey Re- search Center at the University of Michigan, interviewed those employed 20 or more hours per week (thus excluding many part-time workers) in 1972-73. A subset was reinterviewed in 1977, and we use the file consisting of those interviewed in both waves. The QES wage is annual earnings divided by 52 times hours worked per week (weeks worked in the previous year were not available). Both the EEEC and the WDS are probability samples of private, nonagricultural establishn~ents,conducted by the Bureau of Labor Statistics. T h e probability of selection is approximately proportional to employment in the establishment. The WDS excludes supervisory workers. The MWES is a survey of establishments conducted by the Survey Research Center in 1980. In addition to oversampling large establishments, it also oversampled those with minimum-wage workers. Consequently, we use the weights calculated by the center in weighting the MWES wage equations. T h e b1M'ES gives wage distributions (fraction of workers in each of seven intervals) from which we calculated an average wage. The follo\ving sources present detailed information on these data sets: CPS, hlellow (1982); QES, Scherer (1976)and Kwoka (1980);EEEC, U.S. Bureau of Labor Statistics (1982); WDS, Gilroy (1981); and MWES, Converse et al. (1981).
If company sales/(sum of reported establishment sales) was less than one or greater than 100, we deleted the observation as an outlier.
SIZE-WAGE EFFECT 1°33
Census 19771 for establishments and unpublished Snlall Business Ad- ministration data for companies).4
For each data set, size effects are given for the total sample and for various subsamples, defined in terms of broad occupational category and unionization. As the column headed "Other Independent Vari- ables" indicates, we control as much as the data allow for variation in labor quality across different-sized employers.5
Despite the variety of data sets used, table 1 provides consistent support for these conclusions: (1) For the private-sector wage and salary work force as a whole, there is a substantial wage differential associated with establishment size (with company size not controlled for) even in the presence of controls that would be expected to cap- ture much of the cross-employer differences in labor quality: an em- ployee working at a location with ln(emp1oyment)one standard devia- tion (which equals about two) above average can be expected to earn 6-15 percent more than a similar employee at a location with ln(em- ployment) one standard deviation below average. (2) For the same work force, there appears to be a company size-wage effect when establishment size is controlled for and vice versa.6The company size effect is weaker statistically as well as practically. This muj be due to less accurate measurement of company size, although the very indi- rect evidence on this possibility is not clear.' (3) There is clear evi- dence of a size-wage effect in each of the three subgroups of workers. Subtler questions-the relative ranking of the three groups' size-wage effects or the relative importance of establishment and company size for each group-receive different answers with different surveys. (4)
When the category boundaries differed (e.g., QES has categories of 1,000- 1,999 and 2,000 +, while County Business Patterns reports only 1,000+), we assumed a Pareto upper tail in estimating mean employment.
" We use ordinary least squares (OLS) estimation in table 1 and later tables. We also tried correcting for possible (unspecified) heteroskedasticity using White's (1980) pro- cedure. For the CPS equations in table 1, the standard errors of the size variables were less than 2 percent higher than those computed when possible heteroskedasticity was ignored.
"his was previously noted in the CPS data (with categorical size variables) by Mellow (1982) and Oi and Raisian (1985),and by Antos (1981) and Atrostic (1983) for white- collar workers. Dunn (1980, 1984) reported generally similar findings (positive firm size rffects but inconsistent establishment size effects) with continuous size variables in smaller, less representative samples.
One piece of evidence supporting the measurement error conjecture is the fact that, in the WDS data, measuring company size by the logarithm of company sales (which is better measured but probably less appropriate than estimated company em- ployment) leads to appreciably larger company size effects. On the other hand, mea- surement errors are probably least pronounced in the EEEC, where both company employment category and establishment employment are employer reported; yet the size effects are smaller in the EEEC than in the CPS.
A- L' C
F'
1036 JOURNAL OF POLITICAL ECONOMY
When wages are measured by "wages plus fringes per hour worked," size-wage effects are stronger than when the more common "wages per hour paid" are used. Thus the effects of establishment and firm size on fringe benefits and paid vacations and holidays are stronger than their effects on hourly wages.8
Table 1 implies that measured dimensions of labor quality cannot fully explain the size-wage effect. Compared with regressions that control only fbr broad occupation, the additional labor quality vari- ables in table 1 reduce the CPS and QES establishment size and CPS company size effects by roughly one-half.
We also estimated equations similar to those in table 1 for six broad industries" using CPS and EEEC data. The pattern of size differen- tials (particularly the sum of the two size coefficients) was quite similar across broad industries. Moreover, there was no tendency for an in- dustry with larger- or smaller-than-average size effects in one data set to show a similar result in the other one.
Table 1 controls for labor quality by holding constant the worker characteristics that are most obviously related to earnings. An alterna- tive conlplenlentary strategy is to look within very narrowly defined occupatio~ls.We have explored this approach using data from the Area Wage Surveys (AWS) and the Professional, Administrative, Tecllnical, and Clerical Worker Survey (PATC),both conducted by the Bureau of Labor Statistics. The AWS covers 32 cities over the
period 1968-82. The PATC data are based on nationwide surveys for 1965--82. In both data sets, average wages in each occupation and average employment per establishment can be calculated for two size
classes, "large" and "medium.""'
The AWS and PATC data are complementary in that the AWS
includes blue-collar occupations while the PATC provides more white-collar detail." Three conclusions emerge from these by-
'Positive effects of employer size on fringe benefits have been documented previ- ously by Antos (1981),Freeman (1981),and Atrostic (1983).
"he industry groups were mining; manufacturing; transportation and public utilities; trade; finance, insurance, and real estate; and services.
'' We have avoided using "small" to characterize the smaller size classes because really small establishments are not surveyed. In the AWS, large establishments employ 500 or more workers, rvhile medium establishments employ 50-499. In the PATC survey, large establishments employ 2,500 or more workers, while medium establish- ments employ 100-2,499 workers. The lower bound is, in some cases, 100 rather than 50 in the AWS and 50 or 250 rather than 100 in the PATC.
" The AM'S data also include the fraction of office and nonoffice workers covered by collective bargaining in each city-year, but these are not tabulated separately by size class. In analyzing the AM'S data, we assumed that the logarithm of the wage depended on a city-year-specific fixed effect, the logarithm of establishment size, and interactions of size with unionization and time. The linear effects of unionization and time are
SIZE-WAGE EFFECT 1°37
occupation analyses (available from the authors): (1) There is clear evidence of higher wages in larger establishments; the size effects are centered roughly on .05, which is not very different from the esti- mates in table 1 when firm size is not held constant.'"2) The AWS and especially the PATC provide information on different grade levels (corresponding to different levels of responsibility) for white- collar occupations. A striking regularity among the professional, tech- nical, and managerial workers is the tendency for the wage differen- tial to decline with increasing skill level. Whether one interprets this as a true difference in size-wage effects or as a difference in levels of unmeasured skill within grade levels, it seems to be regular enough to warrant attention.13 (3) Both data files show a general pattern of
increasing size differentials between the late 1960s and early 1980s. We also analyzed salary and fringe benefit data for professional and managerial employees of different-sized firms made available by Hay Associates. Here the occupational stratification was based on the Hay rating of individual jobs for compensation purposes (100 points, 200 points, etc.); at each occupational level we regressed the logarithm of compensation on the logarithm of firm enlployment and 27 industry dummies. We again find smaller differentials at higher occupational levels, at least up to the lower managerial ranks, and this pattern persists when fringe benefits and incentive pay are added to salary. At
higher managerial levels, the size differential becomes larger.
Table 2 addresses the question of whether the size-wage differen- tial can be explained in terms of unmeasured dimensions of labor quality whose effect on wages is fixed over time. Such omitted quality dimensions should not be a source of bias when the earnings function is fit using changes in worker characteristics to explain changes in wages (fixed-effect estimates). The size-wage differential observed with cross-sectional data is reduced by 5-45 percent by estimating the
captured in the hxed effect. Taking differences for the two available size classes in any city-year gives
T o make each "size effect" as comparable to the others as possible, we have evaluated the size effect for each occupation at the mean value of unionization and for t = 1982. With the PATC data a similar procedure was followed, except that there is only one (national total) observation for each size class each year, and there are no unionization data.
l 2 T h e blue-collar occupations seem to show slightly larger size effects, but given that the blue-collar occupations in the table are predominantly skilled maintenance work, one should not make a great deal of this difference.
l 3 T h e trend x size interactions tend to be slightly larger in higher-skilled occupa- tions, suggesting that the pattern of size effects declining with grade level was even more pronounced before 1982.
JOURNAL OF POLITICAL ECONOMY TABLE 2
ESTIMATEOSF THE SIZE-WAGEFFECTUSINGLONGITUDINAL DATA:QES, 1973-77 DIFFERENCES
A ln(estab1ishment size)
A union status
A schooling, A experience and
A experience squared, A tenure and A tenure squared
A SMSA, A region (3), A industry (4l),and A occupation (8)
Ln(establishment size)* Union status*
COEFFICIENT (Standard Error)
Yes Yes
Yes ,038 (.007) ,102 (.030)
* These ertltnaLer are frorli "le\el" rnodels thdt are dndlogues to the "chdnge" models excepl that [her ~ricludediirnrnv \aridbles for sex. rdce. and rear, the ertlmdtes werr derlred with the 1973 and 1977 ddtd for the pooled 1973-77 QES sample used In fitting 1he chdnge models (S = 982)
earnings function with fixed individual effects.'" This finding stronglyI' suggests that the size-wage differential cannot be explained solely by appealing to the size-labor quality differential. T h e fixed- effect estimates of the size-wage effect are large in practical terms: if a typical worker went from an establishment with employment one standard deviation below average to an establishment with employ-
'* We also divided the sample into those who changed employers and those who did not. Changes in size in the latter group reflect establishments growing or shrinking and measurement error. The effects of changes in size were larger for the employer- changers than in table 2 and essentially zero for those who did not change employers.
l 5 As Griliches and Hausman (1984) note, fixed-effect estimators are likely to inten- sify the downward bias because of measurement error, and In practice the resulting coefficient estimates are often implausibly small or statistically insignificant. As they suggest, if increasing the period spanned by the two years of data increases the signal1 noise ratio by increasing the amount of "real" change in the independent variables, the 5-year span of the QES is an attractive feature of this file. Another concern is based on the observation that workers choose different-sized employers only if it is advantageous to do so. Consequently, self-selection of size changers could bias the coefficient of the change in employer size reported in table 2. If all changes are voluntary and workers do not care about employer size per se, Freeman's (1981) argument suggests that self- selection will lead us to underestimate the true size effect. Solon (1986) considered a model formally equivalent to the case in which size per se matters. If d is the wage gap between large and small employers and a is the compensating differential required by workers, the change regression underestimates the true wage premium as long as d > a. Since these results assume voluntary job changing, it is worth considering further evidence. If there is a Mills ratio term that belongs in our wage change equation, it
SIZE-WAGE EFFECT '039
ment one standard deviation above average, the employee would en- joy a wage increase of 8-12 percent, about as large as the union-
nonunion differential in these data.'"
Evidence on the Working Conditions Explanation
In a purely competitive labor market, a wage differential not ex- plained by labor quality differences must be due to behind-the-scenes differences in working conditions. Can the size-wage effect be ex- plained in terms of differences in the conditions of work?
While the question is simple, providing a convincing response is not. The reason is that "working conditions" are a very complex phe- nomenon-hard to define and even harder to measure. Since no survey provides an index of the quality of working conditions that would be widely accepted and could be entered into an earnings function that included a size variable, we must conduct a number of less direct but reasonable investigations that, taken together, should permit us to judge the working conditions explanation of the size- wage effect.
Table 3 presents the results of two of these attempts. The first examines the extent to which the size-wage differential is affected by more detailed controls for industry and occupation. More detailed controls should be capturing a greater amount of the variation in working conditions since presumably much of this variation is across industries and occupations.
Our experimentation with industry and occupation controls shows that essentially all the size-wage differential occurs within detailed industries and occupations and thus cannot be explained in terms of a cross-industry or a cross-occupation correlation between establish- ment size and conditions of work. If differences in working condi- tions explain the size-wage differential, there must be sizable partial correlations between establishment size and working conditions within detailed industries and occupations.
To deal with this possibility, our second investigation focused on
should be larger for voluntary job changers. Adding dummy variables for all job changers and for voluntary job (-hangersleft the estimated size coefficient undisturbed. (It increased in the third decimal place.) Similar results were obtained replacing the voluntary change dunlnly with a dummy for those who had lined up a new job before leaving their old job.
16
Evans and Leighton (1987) estimated wage change equations using Sational Lon- gitudinal Survey (Young hlen) data. Their results are broadly similar to ours, if a bit less clear. They find that workers whose firrrt size increases experience a (statistically significant) 5.5 percent wage gain, while those whose firm size decreases suffer a (statis- tically insignificant) 0.6 percent loss. Adding a dummy variable forjob changers makes the wage gain only marginally significant (t = 1.88).
1040
Data Set and Year (Sample Size)
la. hla) CPS, 1979 (13.829)
lc. Same
20. QES, 1973 and 1977 (878)
2b. Same
2r. Same
3a. QES, 197:1-77 longitudinal l ~ l r (439)
3b. Same 3c. Same
JOURNAL OF POLITICAL ECONOMY TABLE 3
ESTIMATEOSF
T H E SIZE- AGE EFFECIW I T H \IARIOUSETSO F F O R N O N W A GWE O R K I N GCOWDITIONS
COWTROLS
Coefhcient of Size X'ariable*
019 (.002) ,015
(.002) .O13 (.002) ,016 (.002) ,015 (.002) ,037 (.006) ,043 (.007)
,044 (.007)
,037 (.OlO)
.033
(.Ol 1)
,028
c 011) ~
Independent L'ariable (Same as Table 1 Except)
S o industrv or occupation dummies
"
dummies (41);"major" census occupation durn- mies (8)
Three-digit census industry dunlnlies (195); detailed census occupation durn- mies (37)
No industry or occupation dunrmies; year dummy Two-digit census industry dummies (41);"major"
census occupation durn-
nlies (8);year dummy Two-digit census industry
dummies (41); "major" census occupation durn- mies (8); year dummy: working conditions vari- ables (I 0)'
(.!iange arlalogur to model 20
(:hang" nnalogur r o r:iodt,l 2!j
C:hange ar1alt.rgue t o niodc! 2~
Size Variable
E C
C E C E E
E
E
B E
~OT-ESw tdl>lc\I~m2ifur~r,f,,rrrl~t~ULI~tl~c\dll,il!mLIXYI!rl rht~.u)dI\x~\urnnldrl.,?d111Ill~\l~t>lc-411 rcsulr, are fur the 1.01 s,~lriplr
* Stantl,ird crrc,rr a l e I,, p.ircnrt~rrcr
1hr ro1A111ic:oild~t~r,n\~\ rl~bldcr\e dcv iibed In thr ICXI
the 1973-77 QES, a longitudinal file that contains information on location size,job conditions, wages, and other factors in both 1973 and 1977. We focused on job conditions that seemed most closely related to issues mentioned in the literature as sensitive to employer size: weekly hours; dummy variables for working on the second or third shift; two variables indicating extent of choice concerning over- time work; variables indicating dangerous or unhealthy conditions on the job and whether the dangerlthreat problem is serious; catchall variables indicating whether more comfortable, pleasant working
SIZE-WAGE EFFECT 1041
conditions are desired; variables indicating whether any of the entire set of job conditions creates a sizable problem; and variables giving commuting time. In order to make clear the impact of adding these variables, all the QES regressions in table 3 are limited to observations for which these variables are available.
Rows 2c and 3c of table 3 present the results of our analysis of the impact of (stated)job conditions on the wage differential associated with location size. T h e findings indicate that the direct information on the conditions of people's jobs collected in the QES can explain very little of the size-wage effect. We also experimented with additional variables intended to measure more elusive working conditions such as pace of work (Oi and Raisian 1985), relationships with co-workers and supervisors, perceived job security, and so forth. Their collective impact on the size coefficient in row 3c was to increase it trivially.
Given the potential problems from measurement errors in the working conditions, which would bias their coefficients downward and reduce their impact on the size coefficient, it is worth asking whether there is any persuasive evidence that working conditions are, in fact, worse in larger employment settings. The lack of such a rela- tionship in the QES data is striking. Of the 42 job characteristic vari- ables included in the regression equation described above, only 21 showed a negative relationship between good characteristics and es- tablishment size, with the other (nonjob characteristic) variables con- trolled for. Of the four significant negative relationships, three re- lated to promotion issues: perceived unfairness in promotions or lack of opportunity to advance. While it is perhaps too much to claim that a difference in working conditions cannot explain the size-wage rela- tionship, our results suggest that it is an unlikely explanation."
Turnover, Tenure, and Wage-Tenure Projiles
The first two rows in table 4 indicate that the quit rate declines with employer size even when the wage rate is held constant. Rows 3 and 4 show that years of tenure with employer (which reflect absence of quits and discharges) grow significantly with employer size, indepen- dent of the size-wage effect; in row 4, a two-standard-deviation differ- ence in size implies a 1.(?-yearor 20 percent differential in employer
Dunn (1980, 1984)tried to assess the disutility of work b) looking at the numher of' dollars workers would pay for (hypothetical) fringe benefits compared to the nunlber of unpaid hours they would work to obtain the same fringes. She found that this disutility rose with firm size in one sample but not in the other; cven in the first sample, the wage premium more than offset the increased disutility. Her results are therefore consistent with our result that, taken together, variations in workirig conditions are at best a partial explanation for the size-wage relationship.
e
0
A
N
Mean of Quit Kate
or Years of Tenurer
.019 (.O08)
,027 (.O 12)
6.34 (8.2)
Coefficient of Size VariableP
Data Set and Year (Sample Size)*
1. Three-digit SIC manufacturing industries, 1958-71 (89)
2. State x two-digit manufacturing industries,
1972 (151)
3. May CPS, 1979 (13,829)
Dependent Variable
In[quit rate/(l - qult rate)]I1
In[quit rate/(l - quit rate)]I1
Tenure with em- ployer
Other Independent Variables
Percentage covered, percentage pro- duction, percentage male, four- firm shipments concentration ratio, In(mean hourly wage)
Percentage union members, index of labor quality, industry (19), region (3). In(mean hourly wage)
Union coverage, sex, race, schooling, experience and its square, SMSA (2),region (3),industry (41), occu- pation (8), In(hourly wage), year dummy
Size Variablei
'044 JOURNAL OF POLITICAL ECONOMY
tenure. These results again suggest that the size-wage differential is not simply due to some nonwage "bad" whose prevalence grows with employer size.
It should be noted that one reason why employees might be less likely to leave a large employer is that there is greater opportunity with a large employer to move from one assignment to another with- out quitting. (Note also, however, that if the worker did not like largeness per se, he or she would have to quit.) If this were all that lay behind the size-tenure relationship, that relationship could not be taken to mean that those who work for large employers remain on their jobs longer because the package of wages and working condi- tions they receive is more attractive than that typically available else- where.
This issue is addressed in the last three rows of table 4. The length of time that QES respondents report working "on ajob like this one" is only weakly related to employer size (row 5);this is consistent with the claim that there is more internal job movement among the em- ployees of larger employers. However, even for those who did not change (three-digit) occupation between the 1973 and 1977 surveys, employer size is significantly negatively related to the probability of changing employers (row 6a),and the relationship is nearly as large as the one for those who did change occupation (row 66). Thus, even among those who remain in the same "job" (as measured by census occupation), those working for large employers are more likely to continue working for them.
Given that table 4's results control for the wage rate, which is posi- tively related to size, we interpret the greater tenure at larger estab- lishments as indicating that large employers offer, if anything, superior working conditions to workers of given quality. An alterna- tive is that they offer packages similar in overall value to those offered by smaller employers but that larger employers' wage profiles are steeper, and so quits are less common.
T o test this alternative, we interacted the logarithm of company size (and sometimes the logarithm of establishment size) with tenure and tenure squared (and sometimes experience and experience squared). We used the May 1979 CPS data, with the same control variables and groups that were used in table 1. The estimated coefficients of the interaction terms were sometimes nontrivial, although as often as not they were statistically insignificant.
T h e results with company size interacted with tenure, tenure squared, experience, and experience squared are representative of the specifications with the largest size-tenure interaction. A two- standard-deviation difference in company size increased log-wage
SIZE-WAGE EFFECT '045
growth per year of tenure by ,003, ,004, and ,012 for the total, white- collar, and nonunion blue-collar samples, and it had no effect for union blue-collar workers. These compare with "average" log-wage growth per year of tenure of .013, ,015, .O 15, and ,004, respectively. However, new workers (those with zero tenure) still receive higher wages if they work for larger firms or establishments, and these dif- ferentials are very similar in magnitude to those reported in table 1.
T h e previously discussed finding that in AWS, PATC, and the Hay Associates data the size premium is larger in the lower grades of (white-collar) occupations than in the higher grades is hard to square with the idea that larger employers offer steeper profiles. Given the mixed results in previous studies (Oi and Raisian 198.5; Pearce 1985), it would be fair to conclude that if large employers do offer steeper wage profiles, the difference is probably not very large."
Evidence on the Union Avoidance Explanation
Since employer size is related to higher wages for union workers (table l), union avoidance efforts cannot be the only reason for size-wage differentials. In table 5, we attempt to determine how important they are for understanding the size-wage relationship for nonunion work-
ers. T h e first row shows by-now-familiar establishment size- and company size-wage effects for nonunion workers. T h e next four rows report analogous coefficients for four groups of workers for whom the threat of unionization is minimal. If union avoidance ef- forts are an important part of the size-wage relationship, that rela- tionship should be much weaker for workers who seem very unlikely to seek unions. We find, however, that the size-wage relationship for these workers is about as strong as that for all nonunion workers. However important union avoidance efforts may be, they are not an important part of the size-wage story.'"
'"A more complicated explanation for lower quit rates among those working for larger employers is that those employers' training is more firm-specific. If small em- ployers offer more general training while large ones offer more specific training, it is possible for the two types of employers' wage-tenure profiles to have similar slopes, but the gap between wages and alternative wages to be growing faster in large firms. U'e cannot test this hypothesis with the data used in this study. It is worth noting, however, that by itself the hypothesis does not explain why those working for large employers earn more initzally.
l9 That managers of larger firms have higher earnings is not surprising given the literature that asks whether managers' salaries depend on sales or profits, especially since an important challenge of this literature is to deal with the high correlation between these ~ariables(see Ciscel and Carroll 1980). Notice, however, that we use a less restrictive definition of manager than these studies (which focus on executives) tend to use.
1046 JOURNAL OF POLITICAL ECONOMY TABLE 3
All nonunion private nonfarm wage and salary workers
hlanagers and administrators
Professional, technical, and kindred workers
Nonunion workers in occupations with union membership percentage 5 5%'
Nonunion workers in industries with union membership percentage 5 5%'
NOTE-Ihe I[,t of independenr \aiiabler ured In each model Ir the Lame a5 the unr uaecl for rou l b of table I
* Standard errors are In parenthese, +-
I'he urrlorl mernbershlp percentdgel used .Ire from Fre~manal:d Zledoff(1SiS)
Evidence on the Product Market Power Explanation
Previous studies have shown that the size-wage relationship survives even when more direct measures of market power (concentration ratios [Weiss 1966; Mellow 19821or industry profits [Pugel 19801)are held constant. One might still wonder whether these are ideal mea- sures of market power. As long as one accepts the premise of these earlier studies-that market power depends on zndust~ycharacteris- tics-industry dummies are a reasonable way of avoiding controver- sies about the correct characteristic(s) to hold constant. Table 3, how- ever, shows that industry dummies down to the three-digit level of detail have no effect on the size-wage relationship.
It is, of course, possible that the products of larger employers are sufficiently differentiated from those of smaller firms in the same indus- t q that the larger producers have less elastic product demand curves and, hence, greater potential profits to share with their workers. To test this conjecture, we analyzed data from the MWES, which asked employers to estimate how their sales would respond to a 10 percent increase in the price of their product, with their competitors' prices held constant. On average, their estimates implied a demand elasticity of - 2.3. With two-digit industry controlled for, the demand was less elastic for multiestablishment firms (by .66, with a standard error of
.18) than for single-establishment firms, but larger establishment size was associated with more elastic demand: dqld ln(estab1ishment size) =
- .31; standard error = .06. Measuring size by establishment size
SIZE-WAGE EFFECT 1°47
alone also suggested more elastic demand at larger establishments (coefficient of - .24, with a standard error of .06).Thus there is little support for the hypothesis that (within two-digit industries) larger employers face less elastic product demands. In any case, adding the estimated demand elasticity had little effect on the size coefficients in table 1 because its effect on wages was small.
A very different way of looking at the ability to pay explanation is to study wage rates of local government employees because for local governments, credible exogenous measures of ability to pay (income or wealth per capita) are available. In a companion paper (Brown and Medoff 1988),we find that controlling for these measures of ability to pay has little effect on the positive relationship between size of local- government employers and the wages they pay.
111. Additional Explanations
A good deal of attention has recently been devoted to formal model- ing of employers' strategies for recruiting workers and for moni- toring and motivating those who are hired. It is not surprising that these models have been used to explain the relationship between employer size and wage rates.
Labor Pools, Worker Sel~ctzon,and Emplojer Size
Weiss and Landau (1984) focus on recruitment and selection strate- gies that minirriize the per unit cost of labor and how these differ for employers that differ in the number of units they employ. Each em- ployer chooses a wage rate to offer to all the workers (in an occupa- tion) it wishes to hire and a minimum level of worker quality. The wage offer determines the quality of the best worker it can hope to attract. The wage offer and the hiring standard must be chosen jointly to minimize labor costs while obtaining the desired number of units of labor.
The key assumption of Weiss and Landau's model is that, as the number of units of labor to be employed increases, the size of the available labor pool does not increase in the same proportion, so the number of applicants per vacancy falls. Consequently, at any given minimum qualification level, the larger employer will be forced to pay higher wages in order to satisfy the greater labor input re- quirement.
If the distribution of worker quality in the firm's area satisfies cer- tain conditions, this mechanism produces a positive relationship be- tween employer size and wage rates. When positive hiring costs are introduced, the model becomes very complicated, although Weiss and
1048 JOURNAL OF POLITICAL ECONOMY
Landau demonstrate a tendency for wages to fall initially and then rise with employer size.
As Weiss and Landau note, their model explains the general ten- dency for wages to rise with employer size.20 If hiring costs per worker are more important at higher skill levels, the relationship between employer size and wages is likely to be weaker at high skill levels, which is consistent with the AWS, PATC, and Hay data pre- sented in Section 11. While it is most natural to think of establzshmer~t size as the relevant variable in their model, they argue that it can explain at least some positive company size effects as well.
The model is too complicated to have derivable predictions about the relationship between enlployer size and quality of worker hired. Thus it is consistent with the positive size-quality relationship we re- ported earlier, but it would be as consistent with the opposite result.
Our reading of the Weiss-Landau model is that it predicts that wages will rise with employer size, eventually, but for reasonably small employers this relationship will be ambiguous. Since "size" here is measured relative to the relevant hiring pool, we expect a weaker relationship between size and wages when the employer is very small relative to that hiring pool. 'Thus we expect the size-wage relationship to be weaker in metropolitan areas or in occupations with national hiring markets. We find little evidence of such patterns."
We also tried to investigate the key premise of the model directly, by analyzing data from the MWES. Employers with minimum-wage workers were asked, "If you were to have an opening of a minimum- wage job now, how many qualified applicants would you get?" Whether or not we controlled for a limited set of demographic char-
20
Weiss and Landau find some evidence in previous work of' a flat or even down- ward-sloping size-wage relationship among relatively small establishments, and (as noted above) their model with positive hiring costs can generate this result. We inves- tigated this possibility with the EEEC and WDS data, which provide continuous rather than catcgorical measures of establishment size. LVe allowed the coefficient of In(en1- ployrnent size) to take on a differerit value at high than at low values of In(estab1ishment size). b.ith the two segments,joined at either 23 or 100 workers. There was no evidence that the additional tern1 mattered in any consistent way.
" When we added an interaction between In(estab1ishment sire) and metropolitan area to the CPS, EE.EC,and M'DS equations in table 1,there was very little difference in establishment sire effects in metropolitan and nonmetropolitan areas. The interaction term was typically right-signcti (i.e., negative) but statistically insignificant and a third or less (typicallv much less) of the In(estab1ishment) coefficient. As regar-ds workers in national markets, table 5 provides no evidence of smaller size premia for-professional, technical, and kindred workers. Even among these workers, however, there are some (especially technicians) ~vhoselabor markets may be more local than national. We therefore examined the effect of' excluding precollege teachers, technicians, and sirni- lar occupations. This I-educed the establishment sire coefhcient to ,007 (.009) but in- creased the compan, sire coefficient to ,017 (.008);so their sum, ,024, was very close to the table 5 value. With establishment size as the only size measure, its coefficient was ,021 (.007).
SIZE-WAGE EFFECT '049
acteristics of the workers in such jobs or for characteristics of the job such as length of workweek and turnover rates, the elasticity of appli- cants per vacancy with respect to establishment size is positive, "small" but statistically significant (typically about .I1 with a t-ratio of 3). Establishments of given size that are part of larger firms also get about
10 percent more applicants per vacancy.22
Holzer, Katz, and Krueger (1988) report an elasticity of (actual)
applicants per opening for the last position filled with respect to estab- lishment size of .16 and mixed results for the elasticity with respect to firm size, holding the wage constant in a broader sample of workers from the Employment Opportunity Pilot Project (EOPP) data. Hol- lenbeck and Mahle (1983, table 2.3) report an elasticity of applicants with respect to employer size of about .23, but with weaker controls for offered wages.
These relatively consistent findings2%an be interpreted in several ways. First, one can accept them at face value and conclude that larger employers have more applicants per vacancy than smaller ones. Sec- ond, they might reflect the influence of unmeasured fringes or work- ing conditions (better at larger employers), though this seems less plausible in the MWES data, which apply to minimum-wage jobs and control explicitly for turnover rates. Third, one can question how successfully the elasticities that are estimated correspond to the rela- tionship that is at issue. If larger employers have several vacancies at the same time, the number of applicants for a hypothetical vacancy (MWES) may overstate the number of applicants per vacancy. Simi- larly, if (as Holzer et al. suggest) large employers are more likely to reconsider applicants for previous openings or convert casual in- quiries into formal applications, the EOPP applicant per vacancy mea- sure may be inflated upward for these employers.
Monitoring
An alternative approach to explaining the size-wage relatior~shipused by some authors is based on the premise that larger employers have
'2 The fact that the question referred to qunl2fied applicants ma) introduce an ele-
ment of ambiguitv to the results if larger enlplo,ers respond to the problem of attr-act-
ing workers b) lowel-ing qualifications. We regard this as a minor problem for two
reasons. First, the average number of "qualihed" applicants pet- vacanc) is h\e; so
responclents obviousl\ used "qualihed" looqel,. I-ather than using it to I-eter to the
qualit) of the one applicant they t!pically hil-e.Second, when we added a ver) limited
set of worker characteristics to coritt-ol for "qualihcations," the results did not change.
21
These results ar-e consistent with Ochs's (1984) experimental stucl~of bu)et-s (alob seat-chers)when the amounts of merchandise available fot-sale (=job vacancies) are known to var) across locations. He finds that hukers choose location5 so that, if anything, the bu\er/nierchandise (= applicarit/\ac.ancv)ratio is greatest at the location with the most stock.
1050 JOURNAL OF POLITICAL ECONOMY
more difficulty monitoring workers. Unlike Oi's (1983) paper, in which larger employers choose high-quality workers to economize on fixed per worker monitoring, these papers argue that large firms' disadvantage in monitoring leads them to monitor less closely. As a result, they are less able to detect the subtler aspects of worker quality (such as effort) and they pay more for workers of given quality.
It is perhaps ironic that a discussion of "recent" work on the subject should begin with Stigler's (1962) classic paper:24
Wage rates and skilled search are substitutes for the em- ployer: the more efficiently he detects workers of superior quality the less he need pay for such quality.
The small company has distinct advantages in the hiring process, so far as judging the quality of workers is con- cerned. The employer can directly observe the performance of the new worker and need not resort to expensive and uncertain rating practices to estimate the workers' perfor- mance. It is well known that wage rates are less in small plants than in large, and the difference reflects at least in part (and perhaps in whole) the lower costs to the small-scale employer of judging quality. A similar result [negative corre- lation with firm size] obtains with respect to dispersion of wages . . . . Men should in general enter smaller companies the greater their ability. [Pp. 102-31
Garen (1985) presents a more formal version of Stigler's model, in which large firms' disadvantage in monitoring leads to different offered-wage schedules, and workers' choice of employers takes this difference into account.
In order to evaluate Stigler's model, we investigated whether it is fully consistent with the results reported in tables 1 and 2 and whether its wage structure predictions are accurate. Because dis- economies in monitoring cannot be measured directly, the empirical tests are necessarily less direct than those used for other hypotheses.
We showed in table 1 that both establishment and company size have independent effects on wages. In the quotation above, Stigler shifts from "company" to "plant" (i.e., establishment) and back to "company," without clearly distinguishing between them. If we re-
"Although Stigler refers to small companies' advantage in the "hiring" process, his analysis really deals with the greater ability to monitor those who already have been hired so that the best workers can be rewarded and, hence, retained. Indeed. large firms have obvious scale advantages in hiring (Hamermesh 1980, p. 387), a larger sample of' observations for detecting the relationship between worker characteristics and productivity. and economies of' scale in studying such relationships.
SIZE-WAGE EFFECT 1051
strict attention to single-establishment firms, larger size is associated with greater monitoring difficulties. But when company and estab- lishment size are not the same, the implications of the monitoring model are less clear, particularly for the eff-ectof establishment size on wages.
In order to explain the observed partzal effect of establishment size on wages, one would have to argue that, if we hold firm size constant but increase establishment size (say, by consolidating the work force into fewer establishments), monitoring of workers has become more difficult. However, it might well be easzer to monitor 1,000 workers in one location than 1,000 workers spread across 10 100-worker loca- tions. Stafford (1980, p. 340) has suggested an alternative possibility: if larger establishments have larger work groups, determining the
productivity of individual workers may be more difficult in larger establishments. Alternatively, it may be that it is really "profit center" size that matters, which might mean that both firm and establishment size are associated with monitoring difficulties.
Stigler also noted that small employers' greater ability to judge worker quality should lead higher-ability workers to select such em- ployers. This argument does not necessarily apply to measures of ability such as years of schooling or years of experience, which are easily observed by employers of all sizes.2i Rather, it refers to subtler abilities that require careful observation (monitoring) to detect. Ga- ren finds that proxies for intelligence are more highly rewarded by smaller employers. However, in table 2 we showed that the coefficient of employer size fell when we moved from OLS to the fixed-effect estimator. Thus those working for small employers appear to have fewer of the subtler virtues that are not captured by the readily ob- served variables in the OLS equation but implicitly held constant in the fixed-effect m~del.'~
The prediction that monitoring difficulties will lead to a relation- ship between employer size and wage structure appears to have re- ceived little subsequent attention because of the difficulty of obtaining the necessary data. One neglected source of such data is the Industry Wage Survey (IMTS)c,onducted by the Bureau of Labor Statistics. T h e
''Garen (1985)argues that the relationship between employer size and the return to schooling is theoretically ambiguous, and he finds that large emplo~ersreward extra years of schooling less than small emplo~ersdo, although this difference is not statisti- cally significant. Stolzenberg (1978) found significantly higher returns to years of schooling in large firms.
'"One might argue that Stigler's hypothesis holds only after one takes account of differences in "skill requirements" ofjobs in large and small firms. But this explariation requires differences in such requirements ulithirl nvo-digit occupations, given the occu- pation d~tnlmiesin our fixed-effect models.
1052 JOURNAL OF POLITICAL ECONOMY
IWS, which surveys individual establishments in selected i n d ~ s t r i e s , ' ~ collects the usual data on establishment characteristics (industry,loca- tion, unionization, and employment) and considerable detail on wage structure for production workers. For example, establishments clas- sify the method(s) of pay they use. Important for our purposes, the IWS distinguishes time-rated from incentive pay systems and, for time-rated systems, permits us to distinguish those in which pay is merit related from standard rate systems in which wages depend only on one's job (and perhaps ~eniority).T'~he main disadvantage of the IWS is that firm size is not recorded.
Stigler's argument is based on the premise that larger eniployers have greater difficulty monitoring the performance of their workers through judgmental rating schemes. One would expect that those
judgmental rating schemes would receive less weight-in the limit, negligible weight by avoiding them altogether-in salary setting by large employers. Such employers should also be more willing to un- dertake the (more costly but more accurate) nonjudglnental eval~~a- tion implicit in a piece-rate ~ystern.''~The evidence in the first two rows of table 6 is quite consistent with these predictions: larger estab- lishments are significantly more likely to use both standard rates and incentive pay (and less likely to use niei-it pay systems).:"'
More generally, if a larger employer's estimate of the productivity of a given worker is less reliable than that of a smaller employer, the larger einployer should place less ~veighton that estimate, and its wage distribution should be relatively compressed (Garen [1985];see Aigner and Cain [I9771 for a similar result in the statistical discrimi- nation literature). Our examination of this prediction, using the two establishment level data sets in table 1, which both have information on the establishment's wage distribution and identify firm size (or, in the case of MWES, multiestablishment firms), is presented in rows 3
" O u r I\SS sample is a set of 10 manufacturing industries previous15 analyzed by Freeman arid Medoff (1984).Their selection criterion was that iridustries have sizable union and nonunion sectors, which should not impart an) particular bias for our purposes.
''
'"or a similar argument, see Goldin (1986)or Lazear (1086).
"
T h e IWS has nine relevant method of pay categories (plus comn~issions~, t-liichis riot relevant to our sample). Of these, four are t)pes of incentive pa). which we grouped together. The five time-rated categories are single rates (paying rver\one in a job the same wage); "range of rates" systems, ~vithprogression through the range hased on seniority, rnerit, or a combination of seniority and merit; and individual determiria- tion. We conlbined "single rates" arid "range of rates: seniority" to form our "standard I-ate" category
Larger estal;lishments probably also have lo\ver per woi ker costs of setting up (and updating) piece-rate svsterns (International Idabc)rOffice 1984), so the latter result by itself could simp]) reHect that advantage rather than a disadvantage in usingjudgmen- tal schemes.
Data Set
(Sample Size) Dependent Variable
1. IWS (3,216) Proportion of production workers paid standard ratest
2. IWS (3,216) Proportion of production workers re- ceiving incentive payt
3a. WDS (1,355)+ Standard deviation of In(hourly earn- ings)
36. Same Same
4a. MWES (978)* Same 4h. Sanie Same
5a. IWS (3,185) Sarne 56. Same Sanie
Other Independent Variables
* Standard rrrors arc ~ r pt a~rrttl~rrrs + .,
-
. .-
Coefficient Size of Size
Variable Variable* E ,022
(.007)
E ,025 (.004)
E ,021 (.003)
E .032 (.004)
C - .011 (.002)
E - .003
(.004) E .005
(.004) 1M - .073
(.012) E ,002
(.OO1 ) E - ,005
(.001)
-.
TABLE 6
Union coverage, sex, SMSA, region (3),industry (21),wage-weighted occuparion index5
Same
Union coverage, sex (2),age (4),
SMSA, region (3),industry (40), pay type (2).average production workweek
Sanie
Union coverage, region (3),industry (60)
Sanie
Union coverage, sex. SMSA, region (3),industry (21)
Sanie as 5a plus ~(occ)~'
Standard tatrs" ~rlcludcaalnglr~rarcsbsrrtrlr and tarngr-c,f-ratrr srstcrna rrl *ha h I,rogrr,rl<,rn tht<,ughthr r.lrnyc la b.iacd on acrltnrltr "Lnuralul*par (~rtccnu\rp,tr hcrornd sorrlr targrt Irbrl <,I <,utput),gtuup Incrrltnc par. and gr<,uph n u r p.!v
* I hrrr r.irnplr a l r o .trr srn.tllrr tharn those I" r.iblc I bec.iurr .tn.rlrs~aot w.tgc di\pcr,~onrcqulrcs dclruon 01 rstahl~rhrrlrntswlth orllr ornr workcr 'IhlrIsrqu.il to In(uagri for tllc rar.ihllahrr~rntif ~tp.!ld r.lch uorkrr thr ~ndu\tr).ivrr.!gr w.i~cfrrr rhar horkrr'\ ~at~of~Ir11l~*
Subscribe to:
Posts (Atom)