Features and Information
In this section,the Journalof EconomicEducationpublishessurveyarticles,
international
and institutionalcomparisons,and analyticalstudieson the economicscurriculum,
instructional
materials,practicesin teaching,andacademic
economics.
WILLIAM WALSTAD,Section Editor
Interview
New
of
Scheduling
Ph.D.
Strategies
Economists
John A. List
Given the importanceof job placement for Ph.D.s, it is surprisingthat economists have not closely examined the factors that affect procuringjob interviews for new Ph.D. economists.' In this study, I investigated those factors
using a data set gathered at the 1997 American Economic Association (AEA)
meetings in New Orleans.My purposewas to increase the informationavailable
to Ph.D. candidates who wish to maximize their postgraduationjob prospects.
In addition, this study may guide undergraduatesand master's candidateswho
seek to pursuea Ph.D. in economics. The results of the findings, however,could
benefit more than job seekers-they may provide academic departmentsand
privateindustrywith a comparativebaseline for making decisions to interview
job candidates.
The job market for new Ph.D.s consists of two submarkets-academic and
business/industry.The following are questionsregardingjob seekersin both submarkets.(1) Do employers in academia seek the same attributesas business/industry? (2) Is there discriminationin the interview decision? (3) Is an MBA
importantin the Ph.D. market?(4) How many more interviews are secured by
candidateswith a finished dissertation?(5) How importantare teaching and research credentials?(6) Are graduatesfrom top-rankedprogramsgiven special
considerationin the job market?(7) Do personal letters of recommendationor
calls from professorsmake a differencein the interviewdecision? (8) How influJohn A. List is an assistant professor of economics at the Universityof Central Florida (e-mail:
[email protected]). The author wishes to thankPeter Kennedy,WilliamGreene, Peter Orazem,
WilliamBecker,Debra Barbezat, StephanKroll, Craig Gallet, Mark Strazicichand five anonymous
refereesfor useful comments.The author also thanksJeff Petry,Josephine Stokes, and JenniferList
for valuable researchassistance. The Universityof CentralFlorida providedfinancial support.
Spring2000
191
ential are recommendationletters from prestigious economists? (9) What is the
marginaleffect of submittinganotherapplication?
A simple theoreticconstructprovides a basis for understandingthe two-step
job-searchprocess carriedout by new Ph.D. economists.2In the first step, thejob
seeker decides whetherto enter one or both submarketsand determinesthe optimal numberof applicationsto submit.The second step reflects the actual decision process regardingacceptanceor rejection of a job offer. Because a natural
prerequisiteto securinga job is an initial interview,my main focus in this study
was to discover the optimaljob search strategiesfor new Ph.D. economists by
determiningapplicantcharacteristicsthat are conducive to obtaininginterviews.
THE DATA
The marketfor beginning Ph.D. economists can be divided into three segments: (1) the preemptivemarketthatcompresses initial interviews,campus visits, andjob offers into the time periodpriorto the annualJanuaryAEA meetings
(e.g., the economistjob marketat smaller,regional meetings such as the Southern Economic Association meetings); (2) the primarymarketthat takes place at
the AEA annualmeetings; and (3) the secondarymarketthat extends from January until late May (Carson and Navarro 1988). Because 70 percent of jobs are
advertisedin the three-monthperiod before the AEA meetings and because a
majorityof initial interviewsfor these positions are scheduled at the AEA meetings, I focused on the primarymarket.
Data were gatheredby personalsurvey from a cohort of first-timejob seekers
at the 1997 AEA meetings in New Orleans.3Participantsin the survey typically
were recruitedby a monitorwho approacheda personattendingthe meetings and
asked if he or she would like to participatein a survey that would take aboutfive
minutes. If the individualagreed, the monitor briefly explained the survey and
usually workedone-on-one with the participantwhile he or she filled out the survey.4A total of 222 participantsgave usable informationon items rangingfrom
maritalstatus to the numberof initial interviewsthey had scheduled.5
The data in Table 1 indicatethat the averagesurvey participanthad scheduled
5.99 academic interviews (denoted Academic), and 1.23 private business/industryinterviewsbefore he or she arrivedin New Orleans.I analyzedonly interviews that were scheduled prior to the meetings and not those obtained during
the meetings. The data indicated that 60 percent of job seekers had 6 or fewer
interviewsin the academicmarket(Table 1). Analyses of the demographicvariables indicatedthatmen represented71 percentof the sample (Male). Of the survey participants,68 percent were white (White),and 66 percent were U.S. citizens. The average age of the participantswas 32.04 (Age); 10 percent of those
surveyedhad a master'sdegree in business administration(MBA);and the average survey participanthad a 3.63 gradepoint average(GPA).6
Those job searcherswho were interviewedalso providedinformationregarding an indicatorof their teaching quality.The indicator,a measure of teaching
quality, was awards or special recognition that the applicants had received for
their classroom performance.Thirty (15 percent) applicantsindicated that they
192
EDUCATION
OFECONOMIC
JOURNAL
had received an awardor obtainedspecial recognitionfor their teaching (Teach.
Award).A job candidate'sresearchcredentialsalso representa potentiallyimportant characteristicon the job market.The averagejob seeker had 0.74 published
articles (# Pubs.) and 0.1 top publications(# TopPubs.). A published article is
considered a top publication if it is published in one of the top 36 economics
journals, as ranked by Scott and Mitias (1996).7 In terms of unpublished
research,the averagesurveyparticipanthad 0.26 papersubmissionsto the top 36
journals (# Top Sub.) and 0.81 submissions (# Sub.) to other refereedjournals.
These magnitudesare largerthan the statisticsreportedby Barbezat(1992).
Ph.D. status may also be importanton the job market.Thirty-onepercentof
those sampled said they had obtainedan economics Ph.D. (Ph.D.) by January2,
1997. This magnitudeis again largerthan Barbezat'ssampledcohort and, combined with the above results, suggests thatjob seekers today are more apt to wait
until they have significantcredentialsin the form of a completed Ph.D. or quality researchbefore they test the job market.
Participantswere also questionedabouttheir institutionof higher education.I
used departmentalrankingsfrom Scott and Mitias to indicateinstitutionalquality. The sample consisted of approximately47 job seekers (21 percent)from each
of the top 3 tiers and 82 (37 percent) from the lowest 140 rankedinstitutions.
Given that the top 30 schools have historicallyproducedalmost half of all economics Ph.D.s (Stephens, Orazem, and Mattila 1994), candidates from lowerrankedschools were oversampled;this may be a negativeconsequenceof on-site
survey administration.
Anotherpotentialindicatorof interviewcounts is the special effort the candidate received from a mentor.Surveyparticipantswere asked whetheran advisor,
or anyone else, had made a phone call or writtena personalletteron their behalf
to potentialemployers. In the academic market(Extra-Academia)45 percentof
those surveyed indicated someone had done this for them, whereas 7 percent
received extrahelp in the business/industrysubmarket(Extra-Bus/Ind).A further
considerationwas reference letters the candidates had received. Because reference lettersfrom prestigiouseconomists may be influentialin the interviewdecision, I used the list of Hall of Fame Economists compiled by Scott and Mitias
(1996) to account partly for the influence of reference letters. Scott and Mitias
ranked50 economists in termsof researchproducedin the top 5 and top 36 journals. Of those surveyed,4 percentreceived referencelettersfrom economists on
one of these lists (HOF Ref.).
Employmentpreferences are another determinantof the interview outcome.
Numerous indicatorsof applicanteffort were gatheredto control for job seeker
preferences. The data indicated that the average job seeker submitted 40.87
applicationsto academic departments[App(Aca.)]and 6.59 applicationsto business/industry [App(Bus.)].8Finally, because the bible of job announcements,
AEA's Job Openingsfor Economists, indicates that opportunitiesfor employment across fields of specializationare highly disparate,I included
primaryfields
of study in the analysis.9Approximately20 percentof those
surveyedhad a field
of specialization in growth/development,resource/environmental,econometrics/statistics,or industrialorganization(Table 1). Fields that were less popular
Spring2000
193
included comparativesystems, history of economic thought, health economics,
and welfare economics.
EMPIRICAL METHODS
To test for optimal job candidate signals, I assumed that the probabilityof
obtainingan interviewwas a function of the job seeker's qualifications.
=
(1)
Prob(yij) J(Xi),
whereyj representsthe numberof initial interviewsjob seeker i obtainedin submarketj (j = academic, business/industry),and X. are the individual'sattributes
(Table 1). A common way to specify this type of discrete probabilityfunction is
TABLE 1
Descriptive Statistics
Interview
Mean
Min.
Academic
5.99
(6.62)
1.23
(2.37)
0
(51 zeros)
0
(140 zeros)
Business/Industry
Exogenous
variable
Male
White
U.S. Citizen
Age
MBA
GPA
Teach.Award
# Pubs.
# Top Pubs.
# Sub.
# TopSub.
Percentile
40th
20th
60th
80th
Max.
0
2
6
10
35
0
0
0
2
20
Exogenous
variable
Mean
Min.
Max.
0.71
(0.46)
0.68
(0.47)
0.66
(0.48)
32.04
(5.21)
0.10
(0.31)
3.63
(0.23)
0.15
(0.35)
0.74
(1.57)
0.10
(0.30)
0.81
(1.2)
0.26
(0.44)
0
1
Ph.D.
0
1
Tier 1
0
1
Tier2
21
57
Tier3
0
1
Tier4
3
4
0
1
0
11
Extra helpAcademia
Extra helpBus/Ind
HOF Ref
0
1
App(Aca.)
0
8
App(Bus.)
0
3
App(Govt.)
Max.
Mean
Min.
0.31
(0.46)
0.23
(0.42)
0.21
(0.41)
0.21
(0.41)
0.35
(0.48)
0.45
(0.50)
0.07
(0.26)
0.04
(0.20)
40.87
(31.64)
6.59
(16.25)
2.53
(4.45)
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
200
0
200
0
50
(Continued)
194
EDUCATION
OFECONOMIC
JOURNAL
TABLE 1---Continued
Field
Mean
Min.
Max.
Comparative
systems
Growthand
development
History
0.05
(0.21)
0.18
(0.39)
0.02
(0.15)
0.14
(0.35)
0.16
(0.37)
0.05
(0.22)
0.14
(0.34)
0.16
(0.37)
0.19
(0.39)
0
1
0
1
0
1
0
1
0
1
Industrial
Organization
International
Trade
Macro.Theory
0
1
MathEcon.
0
1
Monetary
0
1
0
0
1
1
Regional/
Urban
Welfare
International
finance
Labor
Managerial
Micro. theory
Public finance
Resource/
environment
Field
Econometrics/
Statistics
Health
Mean
Min.
Max.
0.18
(0.39)
0.05
(0.23)
0.19
(0.39)
0.16
(0.37)
0.15
(0.36)
0.05
(0.23)
0.09
(0.29)
0.06
(0.24)
0.03
(0.18)
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
Note:Standard
deviationsarein parentheses.
as a Poisson process, which models the expected numberof interviewsas E(y) =
eXb,where X is the regressorvector, and b are the response coefficients of interest.
An importantconcern in the estimation of equation (1) is that the numberof
occurrencesof zero interviews in submarketj exceeds what the Poisson model
would predict. The data in Table 1 indicate that 51 job seekers had zero interviews in the academic submarket(23 percent),and 140 job seekers had zero interviews in the business/industrysubmarket(63 percent).A zero interviewoccurrence in submarketjcould arise from a varietyof reasons,such as carryingout a
limited search, inferiorteaching and researchexperience, or the underlyingdata
generationprocess. Those respondentswho submittedzero applicationsin a particularsubmarketwere consideredto be out of the marketand not includedin the
analysis.After deletions, the sample consisted of n = 211 (academic)and n = 119
(business/industry)observations.Nevertheless, data from these job seekers suggested that there remaineda potentialexcess-zero problem,as 41 (= 20 percent)
and 37 (= 31 percent)job seekers obtainedzero interviews in the academic and
business/industrysubmarkets,respectively.
A technique to account for this potential excess-zero problem is discussed in
Greene (1994). The procedure is a two-step estimation process and has been
given a variety of names, including zero-inflated Poisson (ZIP), zero-altered
Poisson (ZAP), hurdlemodel, or with zero's (WZ) model. When estimatingsuch
models, it is importantto include variablesin stage 1 that signaljob seeker preferences for gainful employment in a submarket.Because considerableinformation is containedin the job seeker's behavior,one importantexplanatoryvariable
Spring2000
195
was included in the first estimationstage-applications submittedto submarket
j. Use of this variable is justified for numerousreasons. For example, if a job
seeker cares little about obtaining employment in a submarket,he or she will
send few applicationsto that submarketbecause of the positive opportunitycost
associated with each application.Hence, in the zero-inflated model, some job
seekers will not clear the initialhurdleand thus will be predictedto have zero initial interviewsin that particularsubmarket.10
RESULTS AND DISCUSSION
Second stage estimation results for the academic and business/industryjob
marketsare shown in Tables 2 and 3.11A preliminaryconcern was the possible
heterogeneityin the interviewdecision across the two submarkets.A likelihood
ratio test indicatedprospectiveemployersacross submarketsconsideredindividual attributesdifferentlywhen makingthe interviewdecision. This result implied
that a unique specificationwas appropriatefor each submarket.12
I employed the ZIP model for estimates for the academic marketbecause the
level of the Vuong statistic (V) in column 6 of Table 2 suggested the zero-inflatTABLE 2
Parameter Estimates for the Academic Market
Zero-InflatedPoisson Model
Independent
variable
Male
White
U.S Citizen
Age
MBA
GPA
Teach.Award
# Pubs.
# TopPubs.
# Sub.
Marginal
effect
-0.86
(.01)
-0.06
(.83)
0.14
(.49)
-0.24
(.01)
-0.03
(.96)
2.50
(.01)
0.83
(.02)
0.25
(.01)
3.68
(.01)
0.29
(.04)
Independent
variable
# Top Sub.
Ph.D.
Tier2
Tier3
Tier4
Extra help
HOF Ref
App(Aca.)
App(Bus.)
App(Govt.)
Marginal
effect
Independent
variable
Marginal
effect
1.13
(.01)
0.64
(.04)
-1.27
(.01)
-1.07
(.01)
-1.60
(.01)
1.37
(.01)
2.92
(.01)
0.07
(.01)
Growthand
development
Publicfinance
-1.44
(.01)
-1.69
(.01)
Diagnostics
Va
Loglike
n
4.54
-679
211
-0.07
(.01)
-0.06
(.05)
estimatesforfieldsof specializaNotes:p ratiosarein parentheses;
p values< .01 areroundedto .01.Coefficient
tionareincludedonlywhentheyaresignificant.
aVrepresents
theVuongstatistic.
196
EDUCATION
OFECONOMIC
JOURNAL
TABLE 3
Parameter Estimates for the Business/Industry Market
StandardPoisson Model
Independent
variable
Male
White
U.S. Citizen
Age
MBA
GPA
Teach.Award
# Pubs.
# TopPubs.
# Sub.
Marginal
effect
1.25
(.01)
0.63
(.31)
-0.49
(.42)
-0.14
(.05)
1.41
(.19)
0.68
(.52)
0.80
(.27)
0.19
(.48)
0.37
(.74)
-0.04
(.88)
Independent
variable
# TopSub.
Ph.D.
Tier2
Tier3
Tier4
Extra help
HOF Ref.
App(Aca.)
App(Bus.)
App(Govt.)
Marginal
effect
Independent
variable
Marginal
effect
0.27
(.01)
1.11
(.10)
-0.57
(.41)
-2.60
(.03)
-1.46
(.08)
3.22
(.02)
-0.70
(.62)
0.01
(.45)
0.02
(.12)
0.08
(.09)
Diagnostics
V
Loglike.
n
1.87
-197
119
estimatesforfieldsof specializaNotes:p ratiosarein parentheses;
p values< .01 areroundedto .01.Coefficient
tionareincludedonlywhentheyaresignificant.
ed model was more appropriatethan the unaugmentedmodel. Coefficient estimates in Table 2 are derivativesof the conditionalmean function with respect to
the attributevector [dE(ylx)/dx= eXbb]measuredat the overall sample means.13
Marginaleffects estimates on the demographicvariables indicated gender was
considered when interviews were scheduled. The model suggested that women
obtained 0.86 more interviews than men, ceteris paribus,which was significant
at the p = .01 level. Although seemingly trivial, given that the mean numberof
academic interviews was 5.75, an expected increase of 0.86 was substantialfor
the average candidate.There was also evidence that older candidatesreceived
fewer academic interviewsthanyoungercandidates--each additionalyear in age
lowered expected interviewcounts by 0.24.
Although these personal traits are important,candidatesmay be more interested in factors they can control. For example, often first-yearjob seekers question the importanceof teaching credentialsand GPA. Estimatesfrom the regression model indicatedthat a candidate'steaching portfolio was highly influential
in the academic market. Coefficient estimates suggested that job seekers who
earned a teaching award also received special recognition on the academicjob
market, as 0.83 more interviews were garneredby award-winningcandidates.
Also, empirical results implied that a one standarddeviation increase (about
Spring2000
197
0.23) in GPA increasedthe expected count of interviewsby more than 0.57.
Coefficientestimateson the researchvariableswere also intuitivelyappealing.
Results implied that academic departmentsdesired candidates who had published in refereedjournals, particularlyjob seekers with publicationsin the top
36 journals. For instance, each peer-reviewedarticle in a highly ratedjournal
yielded approximately3.68 academic interviews. Estimates also suggested that
researchactivity,in itself, was important-submission of researchpapersto refereedjournalssignificantlyincreasedthe expected count of interviews.
Otherimportantconsiderationswere Ph.D. statusand perceivedquality of the
doctoral-grantinginstitution.Coefficientestimates indicate thatPh.D. status was
instrumentalin the interview decision as Ph.D.s received 0.64 more academic
interviewsthanABDs (all but degrees). Estimatesalso suggested thatthere were
some institutionsthat were clearly consideredthe best (top 19) and a groupconsidered the rest (institutionsranked 20 to 240). Ninety-five percent confidence
intervalsindicatedthatinterviewoutcomes were not significantlydifferentacross
schools ranked20 to 240. Nevertheless,estimates suggested a substantialdifference existed between the top schools and the others, as a job candidatefrom a
top-19 institutionreceivedapproximately1 to 1.6 more academicinterviewsthan
a job seeker from a tier 2 to 4 school ranked20 to 240.
Reference letters from prestigious economists also significantly affected the
interviewdecision. Parameterestimates suggested that considerablymore academic interviews were obtained by candidates with a letter of recommendation
writtenby a Hall of Fameeconomist.This finding indicatedthatacademicsearch
committees were highly influenced by recommendationletters from top economists, suggesting candidatescan significantlyimprove their chances on the academic marketif they work with eminent scholars and obtain letters of recommendation from them. Candidates were also buoyed by the extra help they
receivedon the academicjob market.The academicmarketplaceremainsa world
of networking.
Parameterestimates in Table 2 also suggest that the numberof applications
submittedto academiasignificantlyaffected the numberof interviewssecured.A
parameterestimateof 0.07 providedan indicationof the marginaleffect of sending out one more application-at the mean, another submitted application
increasedthe expected numberof interviewsby 0.07. Othercoefficient estimates
suggested that candidatesused submarketsas substitutes.Applications to other
popularsubmarkets,business/industryand government,significantly decreased
the numberof academic interviews secured. Fields of study that were unattractive includedgrowthand developmentand public finance.14
Empiricalresults for the business/industrysubmarket(Table3) were obtained
with the standardPoisson model. I found that the augmentedmodel was unnecessary-the Vuong statistic did not present compelling evidence in favor of the
zero-inflatedmodel (V = 1.87). Marginaleffects in Table 3 suggest that a degree
of discriminationalso exists in the business/industrymarket.At the p < .12 level,
women received 1.25 fewer interviews than men. In addition, older job seekers
interexpectedfewer interviewsthantheiryoungercounterparts,at a rate of 0.14
views per added year. Other coefficient estimates indicated that a finished dis198
EDUCATION
OFECONOMIC
JOURNAL
sertation was a major asset on the private market as ABDs received approximately 1.11 fewer interviewsthan those candidateswith a Ph.D. in hand, which
was significantat the p = .10 level.
Employersin business/industry,like academicians,also separatedschools into
tiers when scheduling interviews.Business/industry,however,had a much larger
range of acceptableschools; they preferredthe top 49 schools over those ranked
50 to 240. Estimates implied that job seekers from the top 19 institutionshad
approximately0.5-2.6 more interviews than comparablecandidatesfrom lower
tier schools. Interestingly,letters of recommendationfrom prestigious economists were not appealingto employersin business/industry.Extrahelp, however,
on behalf of the candidateby a mentoror someone else, was a majorasset on the
private market as parameterestimates indicated that substantiallymore interviews were secured when extra help was received.
Otherparameterestimates indicatedthatthe numberof applicationssubmitted
to the privateindustrysignificantlyaffected the numberof interviewssecuredat
the p = .12 level. A parameterestimate of 0.02 suggested that each submitted
applicationincreasedthe expected numberof interviewsby roughly 0.02.
CONCLUDING REMARKS
Estimationresults suggested a heterogeneityin the interview decision across
submarkets.For example, employers in academia favored candidates who had
quality researchpublications,a completed Ph.D. from a top 19 institution,and
references from established scholars. On the other hand, business/industry
employers sought candidates with completed Ph.D.s from a top 49 institution.
Both submarketsdisplayed a degree of discriminationin the interviewdecision:
Age and gender significantly affected candidates'interviewcounts.
Conclusions from this study provide a basis for possible future research
avenues. Given the finding that a degree of discriminationacross submarkets
exists, it would be interestingto focus on the effects of discriminationon job
seekers' search decisions. For example, do female job searcherspursueemployment opportunitiesin academia ratherthan in private industrybecause of perceived inequities?Although the econometric specifications partiallycontrol for
searcheffort by includingthe numberof applicationsto each submarket,it would
be worthwhile to delve further into this issue by making the search decision
endogenous. Finally, it would be interestingto know how interview decisions
have changed in recent years. In this survey,I covered a period characterized
by
relatively low demandfor new Ph.D.s. An interestingextension would be to link
these results with past findings from strongerjob marketsand comparethe
magnitudes and signs of estimatedcoefficients.
NOTES
1. Two exceptions are the studies of Carson and Navarro(1988; 1985-86
cohort) and Barbezat
(1992; 1988-89 cohort).
2. See Stephens, Orazem,and Mattila(1994) for a derivationof search in a multimarketmodel. I
have also developed a model of multimarketsearch. Furtherinformationcan be obtained
upon
request.
Spring2000
199
3. To maintainconsistencywith priorstudies, I focused on candidateswho were enteringthe Ph.D.
job marketfor the firsttime, thus avoidingany biases associated with includingrepeatsearchers.
4. A copy of the survey instrumentmay be obtainedfrom the author.
5. Gatheringdata onsite has both advantagesand disadvantages.A potential disadvantageis that
only job seekers who have interviewsscheduledin advanceattendthe meetings.Anotherpotential disadvantageof this particularon-site data-generatingprocess is that highly sought-aftercandidatesmay not have the same probabilityof being surveyed as lower echelon job seekers
because of tight interviewschedules.On the otherhand,a potentialadvantageof on-site surveys
is that possible sample selection bias may be mitigated. Previous studies of the economist job
markethave gathereddata by mail surveys, after the primarymarkethas cleared. Because only
successful job seekers may be respondingto mail surveys, the lower end of the job probability
distributionmay not be collected.
6. The surveydataon gender,race/ethnicity,citizenship,and GPA were relativelyclose to averages
from NationalResearchCouncil (1995) data.
7. This rankingsystem had notableshortcomings.For example, the list was constructedwithouta
quantitativebasis, lendingan ad hoc natureto the ranking.For a clear exposition of otherimportant methodologyproblems,see Becker (1997).
8. Because the numberof applicationssubmittedto the governmentsubmarketwas used in the
regressionequations,I includeddescriptivestatistics for the numberof applicationssubmittedto
local, state, and federalgovernment[App(Govt.)].
9. For example, accordingto the "Reportof the Director,"AmericanEconomicReviewPapersand
Proceedings (1997, 498), of the 2,431 availablepositions listed in Job Openingsfor Economists
in 1996, internationaleconomics, macroeconomics,industrialorganization,and economic history were fields cited in 10.4 percent, 11 percent, 8.8 percent, and 0.9 percent,respectively,of
advertisedpositions. Note, however,that a single job is often advertisedfor more thanone field.
10. A final nuance of zero-inflated models is that the changed probabilityinduces a divergence
betweenthe meanand varianceof the distribution,even in the absence of heterogeneity.As such,
zero-inflatedmodels induce overdispersion;the more likely the zero state, the greater is the
overdispersion.Vuong (1989) has proposeda test statistic for nonnestedmodels that has desired
statisticalproperties.The Vuong statistic(V) is directional.If IVJ< 1.96, the test supportsneither
model (e.g., ZIP versusPoisson). If the test statisticis positive and largerthan 1.96, then the zeroinflated model is favored.Very negativevalues supportthe unaugmentedor standardmodels.
11. The numberof interviewseach candidatepotentially secures has an upperbound because time
could representa constraint(the AEA meetings last only four days). This manifestsas an upper
truncationof the dependentvariable(numberof interviews). Both submarketregressionswere
also run, accountingfor truncationat a value equal to 35. Parameterestimates from the truncated regressionsmirrorthose in Tables2 and 3. Nevertheless,to controlfor the possibilitythattime
could be a constraint,or that, for example, interviewsin business/industrywere crowdedout by
interviewsin otherpopularsubmarkets(such as academiaand government),I includedthe number of applicationsto submarketsi and k as exogenous variablesin the regressionfor submarket
j. Estimationresults from the first stage are availablefrom the author.
12. Upon the requestof a reviewer,I used a likelihood ratiotest to test down to a parsimoniousspecification that excluded insignificant variables from the regression. Because these new sets of
regressionsyielded coefficient estimates on the significant variablesvery similar to those in the
full version, I only presentthe full-model results.
13. Non-White and Tier 1 are omitted categories and thereforerepresentbaseline comparisons.
14. I initially included all 18 fields of specializationdummy variablesin the estimatedregression;
only fields with significantcoefficient estimates were includedin the final specification.In addition, I originally estimatedthe model with an indicatorof teaching experience and Black, Hispanic,Asian, andOtherracedummies.Given thatthese variableswere consistentlyinsignificant,
they were excluded from the final specification. Their inclusion did not markedlychange the
results. I also followed this procedurein the Business/Industryspecification.
REFERENCES
Barbezat,D. 1992. The marketfor new Ph.D. economists. Journal of EconomicEducation23 (Summer): 262-76.
Becker,W. 1997. Economics, education,and economists:A view from the United States.Australian
EconomicPapers (Sept.): 5-16.
Carson,R., and P. Navarro,1988. A seller's (& buyer's) guide to the job marketfor beginning academic economists. Journal of Economic Perspectives2 (Spring): 137-48.
200
JOURNALOF ECONOMICEDUCATION
Greene,W. 1994. Accounting for excess zeros and sample selection in Poisson and negativebinomial regressionmodels. Workingpaper# EC-94-10. New YorkUniversity.
NationalResearchCouncil, Office of Scientific EngineeringPersonnel. 1996. Summaryreport1995;
Doctorate recipientsfrom United States universities.Washington,D.C.: NationalAcademy Press.
Reportof the Director. 1997. AmericanEconomic ReviewPapers and Proceedings 87 (2): 497-98.
Scott, L., and P. Mitias. 1996. Trendsin rankingsof economics departmentsin the U.S.: An update.
Economic Inquiry34 (April): 378-400.
Stephens, K., P. Orazem,and P. Mattila. 1994. Job search strategiesand outcomes in the marketfor
new Ph.D. economists. Workingpaper.Iowa State University.
Vuong, Q. 1989. Likelihood ratio tests for model selection and non-nestedhypotheses. Econometrica 57 (2): 307-34.
SOLVE
STUDENTS
ECONOMIC
with CAPSTONE
MYSTERIES
The Nation's High School Economics Course
The CAPSTONE course lets your students solve many
economic mysteries. By looking at the choices people make
and just what influences those choices, students learn that
economic analysis can make sense out of puzzling behavior.
The program focuses sharply on reasoning by posing certain
issues or events as mysteries. For example: Why are young
people, the future of our country, the group with the highest
level of unemployment? Students unravel this mystery by
acting as detectives, seeking and observing economic clues,
then drawing logical conclusions.
Mailcouponto: NATIONAL
ON ECONOMIC
COUNCIL
Dept.
EDUCATIONIMARKETING
NEWYORK,NY 10036
1140 AVENUEOF THEAMERICAS,
Please send the complete CAPSTONEteachingpackage: TeacherResourceManualand
StudentActivitiesBookplus a bonus set of resourcematerials-all in a convenientstorage
box. I understandthe NationalCouncilwillbillme for$119.95, plusshippingand handling.
Position
Name
School
StreetAddress
City,State, Zip
Purchase
School
(must be included)
Order No.:
Telephone
No.:
Send more informationon CAPSTONE.
Spring 2000
Date:
JEE 2000
201