Beruflich Dokumente
Kultur Dokumente
org/stable/27640933 >**"++"$2&QHRFEREFQF&EE2JK
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=sage. 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 service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to Gender and Society.
http://www.jstor.org
Research
Reports
Specialization
as a Missing
Link
1984,
when
Cole
academic
scientists
and
to gender in productivity differences to explain these differ have sought to these factors, In addition factors.
that
extent affects
specialization?can
some Although productivity. has been completely the extent of research neglected specialization a probability careers. in two disciplines academic of academics Using sample and simultaneous data collection, and linguistics), modeling, equation primary ization, finds that the extent less is a critical intervening of research specialization than men and thereby lose out on an important means of
and heretofore unmeasured concept?the the process gender by which help explain areas of special have examined researchers in studies of (sociology the author Women their
variable: increasing
specialize productivity.
Keywords:
gender:
academic
science;
research
productivity;
specialization
receive among the highest of occupa scientists consistently in career attainment are appar tional prestige ratings, large disparities Although in between men and women. And while gender differences ent, especially can be explained of grants, and promotion largely by salary, receipt researchers have had a much harder time explaining research productivity,
NOTE: Funds
This for
research
was
the Advancement
Sociological by the American Linda Grant, Joe Galaskiewicz, and three and Laura anonymous Hunter for reviewers their impec
the Arizona Social Organization Breiger, I thank Jason helped improve the paper greatly. research assistance. cable and extensive & Vol. 20 No. 6, December
GENDER 2006
SOCIETY,
2006
754-780
DOI: 10.1177/0891243206293030
? Sociologists for Women in Society
754
755
in research productivity itself. Only 20 years the large gender difference in research Cole and Zuckerman (1984) referred to gender differences ago, as the "productivity because efforts to account for the productivity puzzle" Keith were at that time unsuccessful. Since then, scholars such as differences et al. (2002), Fox (1983, 1985), Ward and Grant (1995), and others individual- and institutional-level factors that help explain have identified variation in productivity, and Xie and Shauman (1998) were able to explain in regression all variation. But the direct effects typically explored can hide important indirect effects, or mechanisms, are an which of a phenomenon. part of a complete understanding important Elucidating the process by which gender affects productivity rele requires measuring as the nature of scholars' vant but previously unmeasured attributes?such almost research programs?and tistical models. correctly specifying their mediating effects in sta
models
I contend that the nature of scholars' research programs, the specifically extent to which (that is, focused on one or a few sub they are specialized is a relevant but heretofore fields rather than spanning many), unmeasured characteristic that can elucidate how gender affects research productivity. to different degrees, It is likely that men and women scientists specialize and the extent to which one's research program is specialized affects pro some researchers areas of spe rates. Although have examined ductivity and Shenhav 1990; Ward and Grant (Barbezat 1987a; Haberfeld the extent of research specialization has been completely 1995), neglected in studies of academic careers and research productivity in particular. Two in academics could work in the same specialty area but differ dramatically the extent to which their successive papers and projects are related to their previous whereas to the same subfield, could contribute repeatedly could produce articles on other topics as well. This neglect is surprising given the widespread anecdotal evidence that matters. The productivity realized from specializing gains specialization seem intuitive. Specializing allows one to master a literature in a subfield, suc thereby making including the relevant debates and research methods, work: One another cialization
a and related research undertakings easier. Specializing allows to know, and come to be known by, other scholars in the specialty scholar for advancement and publication, area, which espe improves opportunities peer review is not always blind in practice. cially given that double-blind cessive These benefits on investments from specializing may and continue working serve as further incentive to capitalize a in the same subfield(s), reproducing in and productivity. The Center for Advanced Study the link between stages, by specialization promising and pro young recruiting
cycle of specialization the Behavioral Sciences recognizes at least at early career ductivity,
756
GENDER
& SOCIETY
/ December
2006
scholars who have "worked narrowly for 6-8 years to get tenure, [and] are now in a position to think more ambitiously about their work and to take greater risks" (http://www.casbs.org/programs.php?snav=programs_ intellectual level (Blau has been explored at the organizational 1994; Carroll 1985), it has not been integrated into studies of academics' career outcomes. To fill this void, I theoretically motivate, develop, and mea sure the extent of individuals' research specialization and incorporate it into to help explain how it is that men have large and consis statistical models relative to women. The goal of this inves tently higher rates of productivity is to help further delineate the processes gender producing tigation rates and to better understand in publication the productivity differentials the significance of productiv puzzle. Toward this end, I begin by discussing on the gender gap. I then turn to my key theoret literature ity and reviewing extent of research specialization?and the two ical construct?the develop critical links inmy theoretical argument: (1) how gender affects the extent of research affects and specialization research productivity. (2) how the extent of research specialization fellows.html). While specialization
LITERATURE REVIEW
Gender Even and Research Productivity
it is is central to academic success, though scholarly publishing inter understudied 1995), even by scholars (Ward and Grant relatively But rates of aca in colleges and universities. ested in the status of women the large and persistent gender especially are important to study. Publication "makes or rates, careers" (Ward and Grant 1995, 202); perfor breaks individual academic mance and salary raises are evaluations, promotions, funding decisions, on the quality and quantity of publications based (Dean 1989; Ferr?e and is critical relative to men's lower productivity McQuillan 1998). Women's demic publication, in such differences and to study not only because of the size and the persistence of the gap but also because other forms of gender inequality are perpetuated by it (Fox 2001 ; in productivity doc Fox and Stephan 2001). The large gender difference in recent has not disappeared umented by Cole and Zuckerman (1984) and McGinnis 1993; Prpic 1992; Long, Allison, years (Fox 2005; Long to women's to contribute Xie and Shauman and it continues 2002; 1998), to higher academic ranks. in terms of salary and promotion disadvantage of sci in the productivity Because the immense variation "explaining of science" in the sociology research objective entists is a major (Long
Leahey
/ GENDER
DIFFERENCES
IN PRODUCTIVITY
757
to Cole and Zuckerman's have responded 159), social scientists call to explain the productivity (1984) puzzle. They have incorporated and parental factors such as age, marital individual status, cohort, rank, structural factors such as and receipt of research funding as well as more 1992, time avail and current department, department prestige of PhD-granting et al. and disciplinary able for research, (Fox 2005; Keith integration factors were 2002; Prpic 2002; Xie and Shauman 1998). These largely in productivity levels (Xie and in explaining successful gender differences Shauman
tions,
satisfaction
work,
or
of to
of Research most
Specialization
focused on the content previous work on specialization that is, the specialty areas scholars engage in, I highlight of specialization, each scholar's distribution takes by examining the form that specialization areas. To what extent do they specialize, of their of specialty regardless of the extent of research speciali scholars?or rather their research which along the extent to which an individual's located. It assesses programs-are body or diverse. It is easy to imagine a cat work is internally homogeneous of of each scholar's various subfields, with some schol egorical distribution to one ars' research spanning several subfields and others' being confined area(s) zation of specialization? as a continuum I conceive scholars repeatedly engage with the same research topic(s) and tend to write multiple papers on that topic for several years if not their entire careers. I refer to such scholars as specialists. At the other end of the research interests: those who lie scholars with very diverse continuum write papers on a wide variety of topics and rarely publish on the same or two. Some
topic more than once.
of the extent of research there are likely several dimensions Although I focus on the extent to which a scholar repeatedly engages specialization, sur in research on the same substantive topic, for it is the communities research areas that may be critical to producing rounding such substantive specialization's possible: The ulty members specialization are of specialization other dimensions benefits. Certainly, extent of teaching specialization fac could assess whether the extent of service teach the same course(s) repeatedly; a scholar engages in the same kinds whether might capture
758
GENDER
& SOCIETY
/ December
2006
service year after year. Even within the realm of scholarship, of department are possible. For example, one could spe different kinds of specialization in only experimental cialize to a great extent by method (e.g., by engaging to or by theory (e.g., by employing the same theoretical framework work) in sub the extent of specialization studying various topics). I investigate areas of areas because stantive research they correspond best to accepted of Association1?few Sociological expertise as delineated by the American which Moreover, or a single a single method theoretical perspective. embody sur networks and communication it is the invisible colleges or theories? than methods substantive research topics?rather rounding that should patterns Gender be most relevant to scholars' productivity and the gendered of publishing. and Specialization
in differ Previous research has found that men and women specialize and I also expect them to ent areas of research (Grant and Ward 1991), to be a form of pro to different degrees. I consider specializing specialize fessional that begets all sorts of valued outcomes, including pro capital studies have found that women and visibility. Because previous ductivity Vaughn are disadvantaged with respect to other forms of capital (Lin, Ensel, and I also 1981; Paglin and Rufolo 1990; Tarn 1997; Weeden 2002), to be disadvantaged in terms of professional is, capital-that expect women to specialize less than men. Like human, social, and cultural capital, the is likely more form of capital that specialization represents professional to men than to women. available and Productivity I conceive as a resource?essentially of specialization scholars to reap benefits from expect a form of
specializing. capital?I to promote research productivity. I expect specializing Special a scholar to gain in-depth knowledge of a body of literature? thus and key players?and including its central debates, theories, methods, to on that topic easier and more efficient successive should make papers thereby the quantity increasing networks within professional of getting papers accepted peer review is not always of research produced. Moreover, by a subfield, specializing may increase at peer-reviewed journals, given that
write,
blind in practice and reviewers may from specializing in favor of those they know. These benefits be biased on investments serve as further incentive to capitalize and continue may double-blind
759
working
reproducing
a cycle
of specialization
and
on productivity may vary by dis that status attainment patterns are cipline. Previous to be higher in the is known associated with disciplinary consensus, which even across the natural sciences 1982). However, (Hargens and Hagstrom which have been typically understudied and humanities, social sciences resources the processes whereby and Mallard Lamont, 2004), (Guetzkow, specialization research has found are transformed and Gregorio into scholarly productivity 1981). One discipline core (Dogan and Pahre strong losing are likely to vary (Wanner, in the humanities, linguistics, scholars 1989b) from which
touch with the others (phonetics, phone In this field, specialization may mics, morphology, syntax, pragmatics). In contrast, sociology has have a particularly strong effect on productivity. into a large of its subject matter and fragmented the definition expanded number Pahre of poorly connected, 1989b). Specialization of productivity. determinant Overall, and mostly hybrid, be less valued may subfields (Dogan in this field and and less
to our under my investigation yields two primary contributions and in academic work. First, I conceptualize of gender inequality standing construct: an important but heretofore neglected a way to measure develop I incorporate this construct the extent of research specialization. Second, into a model of gender inequality in academic productivity, focusing on its the processes role as an intervening mechanism. by which Understanding and has is critical for policy purposes arises and is maintained inequality a recent American Association been highlighted president Sociological by as researchers in science as well stratification studying (Reskin 2003) these goals, I 1991; Long and Fox 1995). To accomplish from two of academics data on a probability sample a way to empirically measure the extent of research spe fields, develop to simultaneous and specify a process-oriented cialization, equation model new construct. assess the mediating effects of this (Grant and Ward collect extensive
individuals
760
GENDER
& SOCIETY
/ December
2006
data on publications. The sample I have selected retrospective 20 percent sample of tenured and tenure-track sociology study?a at extensive2 research universities?makes guistics faculty members of gender population and research from which
for
this
Research
(National stages leading up to academic employment Council 2001; Preston 2004; Rosser 2004; Sonnert and Holton such as teaching colleges 1995) and to work at less prestigious institutions, and Ward Because the women inmy sample have career pat (Grant 1991). terns that most closely mirror men's, testing for remaining gender differ ences is challenging. research Moreover, by selecting only extensive universities, I control for many resource-based influences research funding, including not only expectations, also time available to devote to research relative to teaching, which has been shown to affect productivity the vast majority of (Fox 1992). Because research takes place at extensive research universities (Levin and Stephan 1989) and the quantity, quality, and form that research takes is rather con sistent across these context on productivity, and travel expenses but
effects particularly specialization stringent. Thus, the I drew my sample is rather selective, and this is par true of the women, who are more likely than men to withdraw from
(Xie and Shauman 1998), they serve as an settings for investigating the impact of research specialization. appropriate = on two disciplines: I focus my investigation (n 196) and lin sociology = in studies of gender stratification (n 222). While many previous guistics on a single discipline academe have focused (Keith et al. 2002; Long 1992; Reskin 2001; studies ification two disciplines because processes of gender strat 1977), I examined have been found to differ across disciplines (Fox and Stephan Levin and Stephan In addition, most previous 1989; Prpic 2002).
of productivity have examined the natural and life sciences (Levin and Stephan and McGinnis 1989; Long, Allison, 1993; Prpic 2002; Reskin to the social 1978), and I am eager to see if the findings are generalizable sciences and the humanities, which have generally been neglected by the of science and knowledge and Mallard Lamont, (Guetzkow, sociology 2004). In these fields, more than in the natural sciences, women have been to a greater extent.3 Compared to other fields such as engi incorporated and linguistics are less sex segregated (Etzkowitz et al. neering, sociology sizes large enough to obtain a high 1994) and therefore provide subsample level of statistical power for examining the gender differences. Although two disciplines also differ in age (sociology is older) and size (sociology is in terms of membership), both sit at the crossroads of several bigger they an investigation and are inherently integrative, making different disciplines of research specialization in these two fields particularly informative.
761
Data
Sources of the data To construct I obtained for my the sampling faculty members' used are culled from secondary analyses frame, take the sample, and contact names and contact information from sites and also from disciplinary guides to
and department Web I also used this informa and membership directories. graduate programs tion to obtain information about where and when the academics earned their doctoral degrees and used first names and various Internet searches, to determine when necessary, each individual's gender. To handle unmea that remains even among the rather homogeneous heterogeneity I control for department prestige research universities, sample of extensive of current and PhD-granting obtained from the National institutions, Research Council and Flattau Maher, 1995). (Goldberger, to construct measures Most of the data needed of research specializa sured tion and research lected from Abstracts productivity
discipline-specific and Linguistics and Language Behavior Abstracts. By entering I accessed academics' all of their refereed journal articles. For names, I collected each article, the keywords used to describe the research, codes (up to two per publi including both the more general classification cation, selected from a total of about 130) and the more detailed major
selected from almost 10,000), both descriptors (up to nine per publication, of which are assigned by trained staff at Cambridge Scientific Abstracts, that manages the umbrella both databases. Because of the organization extensiveness of Sociological Abstracts' and Linguistics and Language Behavior (they index articles from 1,246 and 1,809 than 30 years), they are during a period of more periodicals, respectively, better than other databases such as the Social Science Citation Index and cover only 140 sociology the Arts and Humanities Citation Index, which coverage journals since Information 1970 and needed 1,138 arts and humanities to construct various control journals since 1990.4 variables (e.g., marital was obtained from aca Abstracts'
and parent status and receipt of external funding) I invited all demic CVs and a Web-based survey. During spring 2004, to take a short Web-based academics and submit a copy survey sampled rate was 50 percent, but I was able to obtain of their CV. The response CVs via the Internet for a majority of nonresponders. CVs Importantly, obtained from individuals and indirectly from the Internet are directly to each other and also provide to other data comparable comparable sources of career histories such as disciplinary association publications and Rosenfeld (Dietz et al. 2000; Heinsler 1987).
762
GENDER
& SOCIETY
/ December
2006
Key Measures I assess scholar's the extent research in a taps differentiation From the disci program, using keyword descriptors. or Linguistics databases Abstracts and (Sociological of specialization, which
pline-specific I compiled Behavior for every faculty Abstracts), Language keywords member's multiple articles5 to identify the specialty areas that each journal paper covers (and as noted earlier, the list of specialty areas is different for and linguistics). To construct a measure of the extent of research sociology
I compare the cumulative number of publications with the specialization, to differ cumulative number of unique keyword descriptors. Specifically, entiate scholars who specialize devote a large portion of their research (i.e., program to a small set of specialty areas) from those who branch out (i.e., write successive papers on new topics), I use the ratio of the cumulative to the cumulative number of unique keyword descriptors number of publi cations using data for three hypothetical (see Table 1). Underlined keyword sociologists are those that are new to each scholar's research program (as descriptors defined by a history of journal publications). 1, who first began Sociologist in 1996, has the most diverse research program, covering seven publishing over the course of nine publications. subfields within unique sociology 3, who entered the field in 1984, has the most Sociologist specialized and social research program, covering only two subfields (group processes in his 12 publications sociolo network analysis) since that time. Similarly, two subfields in and profes (1020: occupations gist 2 has only published sions, and 2983: sociology of gender) but is professionally younger and has score than soci and thus he has a lower specialization fewer publications, codes that articles receive at the most two classification ologist 3. Because value of "-1" for schol describe their content, the measure has a minimum = ars who pursue new areas with each publication -1) and a (e.g., [l-(2/l)] value close to "1" for scholars who publish repeatedly on the maximum same topic (e.g., [1-(1/12)] = 0.92 for a scholar with 12 publications and [1 = 0.90 for a scholar with 10 To account for the fact (1/10)] publications). value of specialization that the theoretical maximum depends on number of I assess the robustness of my results to a relative cumulative publications, measure cialization of specialization that indicates the percentile location of one's spe level. score, given his or her productivity scores also depend on how detailed the keyword descrip Specialization tors are (Blau 1977), and for this reason, I construct the measure using key at three levels of specificity. In addition to using the classification (l-[# of unique keywords I demonstrate how I construct / # of publications]). the extent of specialization measure
words
gender of sociology
of 1941: 1941
of
of 1020:
: 1941 family family : 2151: delinquency the 2151: delinquency 2983: the 2151: delinquency juvenile the gender juvenile juvenile of sociology 1636: law 1636: of law sociology law of sociology of law 1636: of sociology 1636: 2148: social work Code 1 sociology sociology Classification sociology sociology
and professions professions occupations 1020: 1020: professions andprofessions occupations 1020: occupations and and family 2983: of gender sociology 0309:group occupations processes group 0309: processes
Stratification in Research Mobility and of Specialization of Extent the Journal Journal the and of Family American Marriage Measure Sociology of the and of Sociological Review American Family Work and Occupations Society Gender and Social Forces Sociology Group Advances in Processes Social Quarterly Psychology
Research Panel A: Publication-Level Information Social Used To Construct Science the Measure Outlet Publication
Marriage Journal Society Gender Society and SocietySociological Law Law Inquiry and Behavior Deviant and Forum Social Forces Sociological
2 Sociologist 20012002200320032004
3 Sociologist 19841985
0.60 0.83
Code group group 1 group group group Classification social social social social social 0309: 0665: 0309: 0665: 0309: 0665: 0665: 0309: 0665:
0309:
Classification NOTE: this that to codes scholar's research program panel new are underlined in A.
Sociological Methods Research and Sociological Social American Psychology American Review of Sociology Quarterly Journal
(CCs)
Sociological Methodology
Social
Leahey
/ GENDER
DIFFERENCES
IN PRODUCTIVITY
765
used in the example above, I also calculate the extent of specializa and also using the broadest tion using the very detailed major descriptors a parent classification code code families, which encompass classification a "social control" family includes the subcodes and several subcodes (e.g., of law" and "penology"). Note that because up to nine major "sociology score derived from descriptors can be applied to an article, the specialization codes value of-8 the maximum these keywords takes a minimum (i.e., l-[9/l]); score 1. Importantly, constructing the specialization still approaches code families allows me to each scholar's distribution of classification using are unevenly account for the possibility distributed that men and women value across have female subfields. fewer This is critical codes if female-dominated possible scholars appear artificially specialized.6 with previ and to achieve comparability To tap research productivity ous studies, I rely on a count of journal articles, arguably the most impor universities. at extensive communication of tant form scholarly but rather a strong to research productivity is not equivalent Publication no single measure is of productivity of it (Fox 1989). Although or universally of ref 1992), quantity (Fox 1983; Long accepted adequate ereed journal articles, perhaps standardized by time, is used most often in the literature (Allison and Long 1973; Fox and 1987; Ferber and Loeb Faver 1977, 1978; Wanner, 2003; Prpic 2002; Reskin 1985; McBrier indicator Lewis, and Grant and Gregorio 1981; Ward 1995; Xie and Shauman of publi dimension also taps the most fundamental 1998). This measure raw article counts are highly cation: frequency 1992). Moreover, (Long status and counts weighted with correlated (Levin by coauthorship or by journal quality (Keith et al. 2002; Levin 1989; Long 1992) Stephan and Stephan 1992; McBrier 1989; Long 2003). Rather than rely on self than male-dominated areas of specialization areas, thereby making
Lewis, reports as others have (Fox and Faver 1985; Prpic 2002; Wanner, and Gregorio 1981; Xie and Shauman 1998), I obtain article counts from I take the databases. To limit right skewness, publicly available electronic counts (Reskin 1977). natural log of raw publication and research of both research specialization my measures Importantly, do not include books. This is largely because books are rarely productivity even if databases. Scientific Abstracts' indexed in Cambridge However, book counts could be obtained from CVs or from the Library of Congress, Scientific lists of keyword descriptors would be different from Cambridge Abstracts' and of specialization impossible), (making a uniform measure with an arti itwould be unclear how tomake a book count commensurate not subject matter (embodied and evidence, cle count. Because method by
766
GENDER
& SOCIETY
/ December
2006
the keywords used here), distinguish books from articles (Clemens et al. the exclusion of books should not bias the measure of research spe 1995), the same cannot be said for the measure of research cialization; however, oriented scholars?a of whom productivity. Qualitatively large percentage are women?tend to publish in book form; thus, my measure their work To ensure that my results are compa undercounts their true productivity. of research pro rable to previous work on this topic, I restrict my measure to articles; in this measure should be the bias however, ductivity considered when interpreting results.
Modeling To
Strategy
on test my hypotheses about the impact of research specialization I use how it may mediate the effect of gender, productivity, especially simultaneous also known as path analysis. This statis equation modeling, tical method is ideal given that (1) the endogenous of interest variables and ultimately article counts) are continuous and (2) I am (specialization interested in not only the direct effects of gender on productivity but the indirect effects that gender might have through specialization (Bollen I estimate is depicted in Figure 1 1989). The ultimate, best-fitting model without error terms and intercorrelated variables for (presented exogenous I use the statistical package AMOS aesthetic purposes). and its full infor some mation maximum likelihood estimation because, when procedure data are missing, this strategy permits the inclusion of all available data is (Anderson 1957) rather than deleting data in a listwise fashion, which the default strategy in most statistical packages. It also bypasses the need to impute data, which is ideal given that sophisticated multiple-imputation are difficult to combine with simultaneous procedures equation modeling.
RESULTS
statistics for my sample confirm findings from previous Descriptive which document stark gender differences in productivity studies, (see Table 2, panel A). Among tenured and tenure-track and lin sociologists the mean number of cumulative articles is universities, guists at extensive men and 9.5 for women. 15.5 for In addition, men have a much larger among men. Even who are statistical 13.5 to women's 8.1, indicating greater variability after removing 18 individuals (8 men and 10 women) outliers in terms of article counts, the stark gender dif ference remains: Men's mean is 16 and women's is 10. This gender dif ference is large and statistically to a /-test for significant according in group means. Although differences differences in these disciplinary deviation: standard
767
FIELD
CAREER AGE
GENDER
RESEARCH PRODUCTIVITY
EVERMARRIED
HAS KIDS
PUBLIC
INST.
DEPT.
PRESTIGE
PHD PRESTIGE
EMP. HISTORY
EXT. FUNDING
Figure
will be examined in the multivariate relationships note here that these relationships remain for both tics but are slightly stronger in sociology.
analyses sociology
768
GENDER
& SOCIETY
/ December
2006
TABLE 2:
Gender
Differences
in Publication
Panel A: Cumulative
Standard deviation Panel B: Extent of specialization Using detailed major descriptors Mean
Median Standard deviation 1.31 1.36
t 7.00
9.10
-2.40
-2.20 -0.60
-2.87
Using classification
0.35
Standard
deviation
0.34
0.34
0.49
0.58 0.50
**
0.43
Standard
deviation
Subsample
NOTE: Significant < .001. tp
size
group differences
255
using
162
one-tailed tests are indicated by *p< .10. **p< .05.
is also a strong bivariate relationship between gender and special score constructed the specialization in the expected direction. Using in the sample tend to with the broad classification codes, I find that women mean score (0.35) less than men (see Table 2, panel B). Women's specialize There ization is statistically significant, (0.40), and this difference that men's cumulative research programs have tended to be more suggesting research pro focused on a restricted number of subfields, whereas women's more breadth. This result holds?and are characterized actually grams by the other two kinds becomes significant?when larger and more statistically of keyword descriptors (the detailed major descriptors and the broadest clas score. sification code families) are used to construct the specialization is lower than men's adjusted for both field relationship between gender and productivity, or linguistics) and experience (career age), serves as a baseline (sociology model for the multivariate 1). It demonstrates analyses (see Table 3, model than fewer articles Women tend to publish what I found descriptively: The men?even I use men in the same field with log of article the natural similar years of experience. the gender coefficient counts, Because of -0.15
0.001
0.03
-0.003* 0.15***
?
Coefficient Standard Standard Error Error ModellCoefficient Model Standard 2 Model Error 3
?
R2 0.32 0.46 0.52
?
TABLE 3:onProductivity: The Unstandardized Mediating Effect ofand Specialization Relationship the Gender between since Years PhD 0.44f 0.003 0.0211 0.003 0.01** 0.00
(1 female) Female if-0.05 -0.15* 0.08 -0.09 0.07 0.07 (10.511 Sociology if0.08 0.07 sociology) 0.50t 0.07
Regression Coefficients Standard Errors and NRC department prestige of rank -0.001 ? 0.002 Extent logged of specialization, 0.37f ? 0.026 0.02 0.35f
prestige
(1? Ever yes) if0.22* married ? 0.12 (1? kids Has yes) if0.06 ? 0.05 (1? Public yes) institution if0.15** ? 0.07 (10.27*** External funds yes) if? ? 0.08
Institutional affiliation
of
rank
NRC
Number
andard
Error Coefficient
0.03 R2?
Coefficient
Standard
Error
0.03 N 418 418 Constant -2.40f ? ? 0.09 0.09 Female if female) 0.14 -2.40f (1 ? -0.44*** 0.14 -0.45***
E:
Dashes
indicate
that
the
variable
was
not
included
in
estimate
was
obtained.
Akaike's Browne-Cudeck information Effects TABLE 3: (continued) on Comparative Incremental specialization Gender3
1113.6 1118.1 criterion 0.83 0.83 index criterion index fit fit tp **p ***p < .01. .05. .10. *p -001. 0.95 0.94
407.4
0.96
One-tailed tests. a.
771
academics publish only about 86 percent (e~?15) of the suggests that women articles that similarly situated men do. Gender, field, and career age com bined explain almost 33 percent of the variation in article counts. as an intervening the extent of research specialization variable Adding to this model demonstrates its relevance (see Table 3, model 2). In this bet ter-fitting model, I find increases specialization at the .001 level and in the expected direction. significant Specialization an additional alone also helps explain 14 percent of the variance in = research productivity (i.e., R2model2 [0.46] 0.14). Adding R2modcl 1 [0.32] to this process also reduces the gender's direct effect on specialization productivity cally to -0.093) and renders it statisti by 36 percent (from -0.146 on productivity, In fact, gender's indirect effect at -0.08) is almost three times larger (standardized through specialization on productivity than gender's at -0.03). direct effect (standardized insignificant. specialize Each productivity: that men more of these than women, and that links is statistically
the extent of research to the gendered is critical specialization as much to specialize as men, of publishing papers: By failing are missing out on an important route to increased productivity. These results hold, and results from previous studies are replicated, I control when for other relevant factors 3, model (see Table 3). Clearly, process women Employment history, marital type (public ing, institutional granting variance departments in productivity, and field. Being and family status, receipt of external fund or private), and prestige of current and PhD an additional 6 percent of the explain together career age, gen above and beyond specialization, older professionally, being at other institutions prior
receiving to one's cur funding, rent job all promote levels. As other studies have found productivity 1987a, 1987b; Cole and Zuckerman (Barbezat 1984), there is no child on productivity, penalty although my measures may not be refined enough to capture this effect; see Fox (2005) for more ideal measures.7 These results are also not sensitive to the exclusion of 18 outliers that were iden tified do not depend on the measure of that the value that specialization takes specialization the spe depends on the specificity of the keyword descriptor. Calculating score using the classification cialization codes or the broad classification Results research used. Recall code families, rather than the specific major descriptors presented in Table 3 does not change the direction or the significance of the coefficients of inter even when I calculate a relative measure of specialization est. Moreover, (in score is compared which each scholar's specialization to the theoretically in the descriptive analyses. obtained from this final model
der, research
772
GENDER
& SOCIETY
/ December
2006
values of specialization for his or her productivity level and the possible resultant percentile value is used), the results for links between gender, spe and productivity remain (results not shown). cialization, The results are also robust to other measures of research productivity still article based) and to alternative model I (albeit specifications. When counts to account for coauthorship cumulative status adjust publication articles by 1/n, where n = total number of authors), the relation (weighting and productivity in Table 3 ships among gender, specialization, presented I weight remain. When for Scientific publications by the Institute journal impact factor (ameasure of visibility or quality of the all results remain except that in model 3, gender's direct effect on journal), maintains its marginal and the indicator for productivity significance, at a public institution becomes and negative. employment significant Information's counts, such journals depends on additional factors yet to be measured, I limit publications to a particular time and writing style. When (2000-2004)?which may be critical when period comparing men and women's shorter careers and thus lower productiv women, given typically between gender, specialization, and pro key relationships ity records?the remain. Using this last alternative measure of productivity in ductivity measure combination with a revised specialization that only relies on key high-impact as creativity words from articles published before the year 2000 also lends support to the of effects that I theorized. In other words, it is unlikely that directionality affects specialization. Even when I allow for a reciprocal effect productivity and productivity in path analysis), (which is possible specialization the direction and significance of the specialization effect do not change. These results also hold for both disciplines under study: sociology and In addition to simply controlling in the models for discipline linguistics. Specialization der differences and other explanatory in raw productivity variables are better able to explain gen in perhaps because publishing
between
a multiple in Table 3, I also specified to see presented group analysis whether the processes of interest differed between the fields, as previous research suggests (Fox and Stephan 2001; Levin and Stephan 1989; Liebert the direc 1976; Prpic 2002; Wanner, Lewis, and Gregorio 1981). Although tion and significance of the coefficients and linking gender, specialization, are the same for both fields, there are some differences in the productivity of effects. magnitude ization in sociology has a stronger effect on the extent of special than in linguistics in lin (-0.62) (-0.36). However, on productivity, as has a slightly guistics, specialization larger impact indicated by both the relative size of the standardized coefficients and the R2 values, which are .46 for sociology and .56 for linguistics. Gender
773
a distinct to introduce dimension of this analysis was an examination extent. However, of gender differences specialization?its in areas of specialization might add to this analysis, especially with respect to disciplinary Do men in different differences. and women specialize extent to which and/or their areas, and could this affect the they specialize This could be the case if some areas (like comparative productivity? The goal of historical work) require intensive historical knowledge and language or other areas lend themselves well to extensive and time-consuming have tended to rely on research. In sociology, we know that women and Rong itative methods (Grant, Ward, 1987) and write on gender skills field qual issues
1993; Lutz 1990) more often than men, and (Grant, Ward, and Bottenfield in primary areas of substantive interest con these differences perhaps and their reduced pro tribute to both their low degrees of specialization the distribution of academics' modal ductivity. Examining specialty areas by gender differences. In reveals interesting and discipline disciplinary areas (phonology two most popular modal and syn women's linguistics, are likely to specialize in In sociology, women tax) are the same as men's. and divorce (code 1941) and sociology of the family, marriage, sociology in social group of education (code 1432), whereas men tend to specialize
of religion (code 0410) and sociology identity and intergroup relations are not only specializing less than in sociology, women (code 1535). Thus, men (which is also the case in linguistics) in differ but also specializing ent areas of study. even as men and women toward differ However, sociologists gravitate ent specialty areas, these areas do not differ in terms of their prestige. I determined each acade of keywords, From the individual distributions a measure of subfield pres and also constructed mic's modal subfield(s) code appeared in the number of times each classification tige by counting and Language the disciplines' (ASR and AJS for sociology, top journals in the year 2003. Results for linguistics) from a i-test suggest that there is no statistically between the prestige of men's and difference significant areas of research. In addition, when I add this variable to women's typical in Table 3, it fails to reach statistical significance the final model presented and does not alter other findings. did not appear in the top journals normally), tige are not distributed indicates whether about 35 percent of subfields Because at all (and thus values of subfield pres I created a dichotomous variable that a scholar's modal area appeared in a top journal in 2003.
test of independence this dichotomous variable and A chi-square between areas of have equally prestigious confirm that men and women gender = and this finding remains when I conduct the test .174), (p specialization separately for each discipline.
774
GENDER
& SOCIETY
/ December
2006
DISCUSSION
I have found that a heretofore neglected aspect of scholars' a crit extent to which programs?the they are specialized?plays ical role in not only explaining variation in productivity levels but also help lower rates of productivity ing us understand why women have consistently In this article,
research
than men. Having a specialized research program (i.e., writing papers in the same specialty area repeatedly) promotes productivity, but women tend to on a means of out have less specialized research programs, thereby missing rel effect is large (especially achieving higher productivity. This mediating ative to effects of more variables used to account for typical explanatory in productivity), and robust to gender differentials statistically significant, of specialization and productivity. alternative measures of this process will be necessary to understand why Further decomposition women tend to specialize less than men even when they conduct research in at work the same areas, as was the case in linguistics. Perhaps the phenomenon et al. 2003) operates here, but only for men: While expertise, men specialize because they think a diversified research program indicates a failure to excel in any one area, whereas women diversify because they think it indicates scholarly breadth. In other words, women may think that diversification will broaden their profes in the film industry (Zuckerman attempting to demonstrate sional identity, whereas men may fear itwill sully theirs. Men's and women's different professional networks and collaboration strategies may also be rele vant. Men's wider and more diverse professional networks (Burt 1998; Kanter 1977) may allow them to find collaborators whose interests overlap their own, to reinforce their expertise in one or a their research collaborations allowing
few specialty areas, whereas women's smaller and more homogeneous pro
and Moody 1991; Renzulli, Aldrich, (Grant and Ward 2000) require them to branch out to other substantive areas if they want to col laborate, resulting inmore diversified research programs. I found the processes mediates gender differ by which specialization ences in productivity to be roughly the same in two fields typically of science: and linguistics neglected by the sociology sociology and Mallard Lamont, (Guetzkow, 2004). However, gender effects were more pronounced core areas in sociology?a field with many ill-defined in linguistics. Linguistics has only a few (Dogan and Pahre 1989a)?than core each areas (semantics, of which requires and phonetics), syntax, morphology, phonology, a very different knowledge base (e.g., anatomy, grammar, and/or experimental psychology, neurology, a great degree of specialization for likely encourages normatively high degree of specialization in
fessional
networks
biology,
Leahey
/ GENDER
DIFFERENCES
IN PRODUCTIVITY
775
than comparable of linguistics has a larger effect on productivity there is more where in the field of sociology, of specialization
among subfields (Leahey 2006). and research pro of specialization light on the processes other domains of spe future scholars might consider examining ductivity, in terms of subject I focused on the extent of specialization cialization. or theo matter (Turner and Turner 1990), not in terms of methodological cross-fertilization To shed more typically special approach, even though we know that academics as well as substantive and method (Burt 2004). Even topic theory are less specialized substantively, perhaps they specialize though women it is cer more in terms of methodological and theoretical approach. While and the the case that both kinds of keywords (the major descriptors tainly retical ize by codes) include terms that pertain to methodology (e.g., log interaction and theoretical perspective linear modeling) (e.g., symbolic are not distinguished in the current analysis. ism), these dimensions but its effect on other aids productivity, I found that specialization itmay be breadth rather career outcomes remains to be seen. For example, classification in the field, external job than depth of research that promotes visibility to open-ended and salary levels. Responses ques offers, job satisfaction, effects may that specialization's tions from theWeb-based survey suggest on the career outcome under study. It may, for example, vary depending "I think that specialization have negative effects on institutional mobility: to to some extent, and may have kept me from moving was detrimental me as being too another department. may have perceived Departments narrow in focus and even a large department might not have wanted me if in the field" (woman sociologist). working limit the "broad appeal" of one's work and sub may Specialization areas that I have chosen to work on "The particular sequent visibility: in the sense that the questions have helped me reach my current position, they already had also I ask in my research seem to speak to interests of a relatively broad soci audience. For example, when I was on the job market my PhD ological rea for different to interest people with different seemed subspecialties in more than one area has sons. I believe the fact that I have done work Whether an advantage up to this point" (woman sociologist). valued out in research?or any kind of work?promotes specializing is a question comes such as mobility, and even job satisfaction visibility, also been study poses and future studies will ideally address. women spe strong, significant, and robust results of this study?that enhances productivity?are cialize less than men and that specialization just to a theory of a beginning. results can contribute Before these empirical such as policy in work and to more practical endeavors, specialization that the current The someone
776
GENDER
& SOCIETY
/ December
2006
making and career planning, they must be replicated on other samples from other fields, at different times, with different outcome variables?including more encompassing measures of research productivity than the ones I used as defined by here. Does also affect research productivity specialization books, and if so, is the relationship positive, as we found here? Are women's more akin to men's levels of specialization levels when books are used to assess specialization? Do the current results hold up when the analysis shifts or even other professions? to other disciplines These kinds of questions will motivate future research and contribute to the development of a ideally broad theory of how specialization affects gendered career patterns.
NOTES
1. See the list of sections (http://www.asanet.org/page.ww?section=Sections&
name=Overview)
devised
the Carnegie that award
list of specialty
Foundation 50 or more
areas
currently
(http://www2
uses to refer per year
those
doctoral
degrees
in at least 15 disciplines. 3. This is evident in an increasingly equitable gender distribution of recent PhDs over time: In 1973, less than 15 percent of the professionally young (PhD
within ioral with where the past sciences much in 10 years), was lower 1995, women; levels of full-time, this academic increased in 20 labor to 45 the force in the by sciences social This and behav percent 1995. and contrasts
women
natural
engineering, respectively
percent can be
and
11 percent,
(National Research
4. Serial lists and factsheets/socioabs-set-c.php
Council
2001).
schemes found (for Abstracts) Sociological and Language Behavior (for Linguistics between coauthored and single-authored one worked others because whether have, others research not should not diminish or enhance at http://www.csa.com/ and http://www.csa Abstracts). pieces, on a as
classification
.com/factsheets/llba-set-c.php 5. I do not distinguish Wagner-Dobler or in collaboration alone nence of that the subfield surface, (1997) and
6. On
appear
which women are disproportionately represented include 2900 (gender), with two child codes, but also 1900 (family), which has six child codes. Sociological areas inwhich men are disproportionately represented include 1200 (urban sociology), with just one child code, but also 0100 (methods), which has five child codes.
7. Tests did not reach to examine statistical linguists whether significance. receive English their or doctoral training language, in disciplines and cognitive as diverse science, as and another gender interacted with marital or parental status
psychology, discourage
diversification.
777
REFERENCES
Allison, tists. Anderson, distribution P. D., American T. W. and J. S. Long. Sociological 1957. when Maximum some 1987. Review Interuniversity 52:643-52. mobility of academic scien
likelihood are
normal
observations
the American
Statistical Association
Barbezat, ket. -. demic Bayer, A. Debra Journal 1987b. A. 1987a. of Human
52:200-203.
Salary Resources differentials 22:422-28. or sex discrimination? and Policy Evidence Review from 6:69-84. reward system. the aca by in the academic labor mar
Research Sex
E.,
S. Astin.
differentials
in the academic
and heterogeneity:
of academic
A primitive
work. New
theory of social
Brunswick, NJ:
Press.
with
latent
variables.
New
York:
Wiley.
Burt, -. Ronald. 2004. 1998. The gender holes capital. ideas. Rationality American and Journal Society 10:5-46. Structural of Sociology of niche 90:1262-83. 1995. Careers Journal in of and in width
110:349-99.
Carroll, Glenn R. 1985. Concentration specialization: Journal and D. reputation. Dynamics of Sociology Okamoto. American in populations of organizations. American E. S., W. W. K. Mcllwaine, Clemens, Powell, print: Books, journals, H. Zuckerman. of publication and scholarly 1984. The of men
Sociology
Cole, J. R., change Motivation
101:433-94.
and in patterns and productivity and women puzzle: scientists. Persistence Advances
Achievement
2:217-58.
Dean, Dwight
review Dietz, and J. S.,
G.
some
1989. Structural
Bozeman, the
constraints
dilemma: A
Using the An
American
career
scientists
Scientometrics 1989a.
and R. Studies
1989b.
Development International
Science
41:457-70. H., of A., C. Kemelgor, mass J.W. among Loeb. male Neuschatz, B. Uzzi, and J. Alonzo. 266:51-54. and of sex perceptions Journal of Sociology 1994. The
critical and
in science. Performance,
discrimination
female
faculty.
78:995-1002.
778
GENDER
& SOCIETY
/ December
2006
Ferr?e,
M.
M.,
and
J. McQuillan.
1998.
Gender-based
pay
gaps:
Methodological
and policy issues in university salary studies. Gender & Society 12:7-39. Fox, Mary Frank. 1983. Publication productivity among scientists: A critical review. Social Studies of Science 13:285-305.
-. 1985. Location, sex-typing, and salary among academics. Work and
Occupations
-. process. -. 1989.
12 (2): 186-205.
peer review, and the publication versus
1992.
and publication
productivity:
Mutuality and
competition
-. 2001.
in academia. Sociology
Women, science, and
of Education
academia:
65:293-305.
education careers.
Graduate
15:654-66.
Social family Studies 1985. characteristics, of Science Men, and 35:131-50. and publication Quarterly young scientists: productivity: 26:537-49. Preferences, publication productivity
C. A.
social P. E.
academics. 2001.
prospects, and realities by gender and field. Social Studies of Science 31:109-22.
Goldberger, grams Grant, L., M. L., B. A. Maher, States. and P. E. Flattau. DC: and 1995. National Research-doctorate Academy in sociology. Press. Gender & pro in the United and K. Washington, 1991. Gender
B. Ward.
publishing
Society
Grant, L.,
5:207-23.
K. B. Ward, and D. Bottenfield. era as reflected 1993. Women's in journals. Is there in the pre-WWII Meetings. and X. B. Ward, methods in Miami, research sociological FL: American
Grant,
L. Rong.
1987. research?
an association
between
gen Review
der
American
Sociological
52:856-62. Guetzkow, J., M. Lamont, 2004. What is originality and Blacks in the human
Sociological
women
Review 69:190-212.
closing the gap?
science during in American Salary discrimination Industrial and Labor Relations Review 44:38-82.
Hargens, status Heinsler, L. L., and W. O. Hagstrom. 1982. Scientific attainment J. M., Sociology patterns. and R. A. Rosenfeld. of Education 1987. Charting women
the
consensus
1970s
and
1980s.
and
academic
source make
Kanter, Basic Keith, tific B., Rosabeth Books.
a difference?
Moss. 1977.
17:135-44.
corporation. The context and the New York:
J. S. Layne, on 2006.
N. Sex
Babchuk, status,
and K.
Johnson.
2002.
of
scien of
achievement:
Leahey,
publication Erin.
scholarship Conceptual
timing
Measuring
sci
entific
manuscript.
779
Levin,
S. G.,
and
P. E.
scientists. Liebert,
Research J.
1989.
Age
and
research
productivity
of
academic
Education
Roland
Productivity, J. C. Vaughn. in
American
Lin, N., ties: W.
Journal of Sociology
M. Ensel, and factors
82:664-73.
1981. Social status strength American of occupational sex differences
Structural
Sociological
Long, J. Scott. Forces
Review 46:393-405.
1992. Measures of in scientific productivity. Social
J. S., P. Allison, and R. McGinnis. Long, careers: Sex differences and the effects
of productivity. careers:
Review
Long, J. S.,
58:703-22.
and M. F Fox. 1995. Scientific Universalism particular anthro
21:45-71.
of women's writings in sociocultural
17:611-27. Gender case of and career dynamics law academia. Social within Forces a segmented 81:1201-66.
Branch.
professional
labor market:
National
in the
Research
careers
Council.
of doctoral
Washington,
Paglin, M., choice, 8:123-44. Preston, careers. Prpic, Anne New and
Press.
capital, of Labor exits occupational Economics differences.
and A. M.
Heterogeneous
E.
2004.
science:
Occupational
from in
scientific science.
York: 2002.
Russell
Katarina.
and
productivity
differentials
Scientometrics Renzulli, works, Reskin, ence. -. ence. -. American Rosser, the Sonnert, sion. Tarn, Sue struggle G, New L. A., and Barbara. American 1978. American 2003. H.
55:27-58. Aldrich, and J. Moody. matters: 2000. Family outcomes. Social Forces 79:523-46. and productivity Review 42:491-504. sex, and the reward Gender, net
entrepreneurial 1977.
Scientific
structure
of
sci
location
in the
institution
of
sci
68:1-21. glass York: Who ceiling: Routledge. succeeds in science? The gender dimen Press. University and occupational Academic
1995. NJ:
Rutgers
Tony. S.
gender
inequality An
in
the
102:1652-92. of Sociology The 1990. science: impossible Newbury Park, CA: Sage.
institutional
of American
780
GENDER
& SOCIETY
/ December
2006
Roland. A L.
1997.
quantitative
of Social
scientific Studies
S. Lewis,
I. Gregorio. sciences,
1981.
Research sciences,
academia:
comparative
social
ties. Sociology
Ward, K. B., and
of Education
L. Grant.
54:238-53.
Gender and academic publishing. than American others? Higher Social of New
Education:
Weeden, closure Kim and
Handbook
A. 2002. earnings
11:172-212.
pay more States. Journal
Why
Sociology
Xie, -. MA: Zuckerman, tities or Y, evidence 2003. Harvard
108:55-101.
Shauman. an old Sex difference in research Review and about puzzle. in science: Press. K. Ukanwa, in and the J. von Rittmann. feature-film labor 2003. market. Robust iden American Career Sociological processes productivity: 63:847-70.
and K. A.
Women
outcomes.
Cambridge,
nonentities?
Typecasting
American
Journal of Sociology
Erin Arizona
Leahey whose
is an
assistant
of
sociology been
at and
the
of of how
include recently,
methodology,
the sociology
practice.
academics innovation.