Professional Documents
Culture Documents
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=asa.
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.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.
http://www.jstor.org
Replyto Jann
MarketPremium,SocialProcess,
and Statisticism
Yu Xie Xiaogang Wu
Universityof Michigan Hong Kong Universityof Science & Technology
To be fair,Jannshould not be singled out for sectorto the statesector.2In 1978, at thebegin-
falling into the trap of "statisticism"because ning of the Chinese economic reform, 1,197
such practice is so widespreadin currentsoci- respondentsworkedin the statesector.By 1987,
ology that it often makes quantitativeresearch 11 percenthadmadethe transitionintothemar-
unappealing.Jann'scommentillustratesa com- ket sector (d = 1) and are called "earlybirds."
mon temptation among quantitativesociolo- Among the remaining1,068 workersin the state
gists: reliance on canned statisticaltests rather sector and the 522 new entrantswho startedin
than substantiveknowledge.Thus, we take this the sector between 1978 and 1987, 16 percent
opportunityto drawa general lesson for all of made the transitioninto the marketsector(d =
us: only when combined with a substantive 2) and are called "laterentrants."3 The remain-
understandingof the social processes involved ing 1,337 respondentsare called "stayers."
can statistical methods result in fruitful The Wu and Xie (2003) article was partly
research. responsible for causing Jann's confusion,
At a fundamentallevel, Wu andXie's (2003) because his reanalysisis modeled afterWuand
studywas descriptive.We emphasizedthis point Xie's (2003) regressionsof logged earningson
throughoutthe articleby alertingreadersto the education, sector, and their interactions.
dangerof aggregationwhen workersin a single Extensive discussion in Wu and Xie (2003) on
group are in fact heterogeneous.Although we the differentialreturnsto education by sector
questioned the prevailingwisdom that marke- makesit appearas if educationis the causalfac-
tizationper se "caused"the educationreturnsto tor, with sector as a covariate. However, our
be higher, we never intended our statistical real researchquestion concernedearningsdif-
analysis to be more than a descriptiveexercise. ferences by sector, with education as a con-
In such an exercise, formal statisticaltests can founder.Let us revisit a passage (Wu and Xie
be informativewhen there is sufficient statisti- 2003:430) cited and emphasizedby Jann.
cal power. In the absence of strong statistical
Thecrucialdifferencebetweenthetwohypothe-
information,substantiveknowledgeshouldpre- of earlybirds.In Hypothesis
ses is the treatment
vail. 1, earlybirdsare groupedwithinlaterentrants
In his comment,Janntreatsthe three groups becausetheysharethe commonfeatureof being
underdiscussion-stayers, earlybirds,and later in the marketsector.... In Hypothesis2, early
entrants-as thoughthey were symmetric,like birdsare groupedwith stayersbecausethe two
those of an experimentaldesign. In doing so, he typesof workers wereapproaching a convergence,
borrows the language of multiple-groupcom- against which later entrantswere selectively
recruitedintothemarketsector.
parisons commonly used in analysis of vari-
ance (ANOVA) associated with experimental Jannassumesthatthephrase"aregroupedwith"
designs. However, because we were dealing here means "sharethe same educationcoeffi-
with observationaldata,ourconcernswere with cient with."This is incorrect,as shownin Figure
the between-group and within-group hetero- 1 of Wu and Xie (2003:431). It is possible, for
geneity generatedby social processes.The earn- example, thatearlybirds and laterentrantsdif-
ings regimes for the threegroupsresultedfrom fer in earningsand educationdistributionsbut
a cumulative historical process that clearly is have the same educationreturnsas stayers,yet
asymmetrical (Figure 1) and thus should be pooling early birds and later entrantstogether
treated as such in an analysis. Comparingthe still could yield higher returns to education
three groupsas if they were threeexperimental
conditionsis bothinappropriate andmisleading.
Figure 1 presents a schematic flow chart of 2Thatis, we excludeda smallnumberof"market
the respondenttypes in the 1996 survey of Life
Histories and Social Change in Contemporary losers"fromFigure1 becauseof the group'ssmall
size (WuandXie 2003).Webasedtheclassification
China,the dataused by WuandXie (2003). The onWuandXie'scomprehensive measureof themar-
y-axis representsthe employmentsector (state ketsector.
vs. market),andthe x-axis representshistorical 3 Here275 workers,including82 "laterentrants"
time. We make the convenientassumptionthat in WuandXie's(2003)originalarticlewhoentered
the marketsector is an absorbingstate so that the laborforceafter1987 weredroppedfromthe
there is no reverse transitionfrom the market analysis.
AND REPLY 867
COMMENT
NewEntrants to the
StateSector(522)
o
Experienced Stayers Stayers Stayers
a Workers(1,197) (1,068) (1,590) (1,337)
LaterEntrants
S(253)
EarlyBirds EarlyBirds
(129) (129)
amongworkersin the marketsectorthanamong transitionby the end of the study (i.e., a stay-
stayersin the state sector. er).Note thatfor a personwho has madean early
To advancethis inquiry,let us reconceptual- transition(d = 1), the counterfactualoutcome
ize the substantiveproblemwith explicit coun- shouldfollow the principleof "forward-looking
terfactualsin the language of causal inference sequential expectation"(Brandand Xie 2005),
(Heckman2005; Holland 1986; Manski 1995; a combination of a late transition(d = 2) and
Winship and Morgan 1999). Suppose we are staying (d = oo).We thus define the average
interestedin the causalimpactof the entryto the causal effect for the first question as
marketsector on (potential) futureearningsin
1996. Conceptually,there are two causal ques- - = E(Yd=) -
E(ydl) E(yd>l)
tions in this setup: (1) what is the effect of an + E(yd=m)(1- P2)].
[E(Yd=2)P2
earlytransition?(i.e., d = 1) and (2) what is the
effect of a late transition?(i.e., d= 2). Of course, Note thatthe transitionprobabilitiesare condi-
these two questions are inherentlyasymmetric. tional so that P2 = P(d = 2 Id > 1).
The second is sensible only for those workers For the second question, the comparison is
who did not experience an early transition, simpler,involving two regime-specific means:
whereas the first involves the counterfactual
E(Yd=2)- E(Yd>2)= E(Yd=2)- E(Yd=o) (2)
comparisonbetweenthose who experiencedan
early transitionand those who did not, regard- It is neverpossibleto computequantitiesdefined
less whathappenedto them later.To borrowthe by equations 1 and 2 because we observe only
notationfor causal inferencewith time-varying one of the three potential outcomes for each
treatments(BrandandXie 2005), let Yi denote worker. To infer causality, it is necessary to
the ith person'spotentialoutcome if the person introduce the ignorability assumption, which
has madea transitionat time d (d = 1, 2, oo),with mustbe takenas provisionalbecauseit is unlike-
d = oo denoting thatthe person has not made a ly to hold in reality.The ignorabilityassumption
868 AMERICAN
SOCIOLOGICAL
REVIEW
900 -
* 6.07 * Stratum-specific -Linear fit (HLM)
700 -
500 -
Y = 645.2 - 94.3 X STRATUM_RANK
'o 300 -
o 500
t 3.33
0.31
,
S100-.9
-100-
* -0.65
1 2 3 4 5 6 7 8
PropensityScore Stratum