well as comparisons with receiver operator characteristic curves
based on threshold values (see electronic supplementary
material, SI [30]). Ambiguous subjects in which the duplicate
samples gave conflicting results wer e re-run. Results from anti-
body testing were linked to each s tudent’s indicated major, for
which we performed comparisons both between Business-related
majors and non-Business majors as well as wi thin bu siness -
related maj ors by com paring t he ‘manag ement an d entrepre -
neurship’ emphasis with all other business emphases
(accounting, finance, marketing) and specifically to accounting,
which i s consid ered the m ost risk- avoidant em phasis consid er-
ing its hi gh placeme nt rate ( perce nt of stude nts who rece ive job
offers in their selected career field). Our ra tionale was t hat
Business majors would be more i nterested i n entrepren eursh ip
than other majors and within the Business major, students who
study ‘ managem ent an d ent repreneurs hip’ would be more
inclined to wards entr epren eurship than othe r empha ses, su ch
as accounting. The total sample size of subjects with useable
inform ation on inf ection st atus and field of study was 1293
(electronic supplement ary material, table S1).
(b) Study 2: infection patterns in business professionals
To understand p atterns of inf ection amo ng professio nal entre-
preneurs, we coll ected d ata from 197 in dividual s attendin g
entrepreneu rship event s. All indi viduals at th e locations were
invited to participate by providing a saliva sample, as outlined
above. Particip ants were given a col our-code d vial based o n
their sex and wh ether or not they h ad ever started th eir own
company, and s amples were proc essed as des cribed above.
For both Stud y 1 and Stud y 2, we used gene ralized line ar
models with a binomially distributed response and the comp-
lementary log – log link function to evaluate the influence of
T. go nd ii inf ection status ( antibody -negative versus a ntibody-
positive) o n major fiel d of study (Bus iness majo r versus
non-Business majors, and ‘management and entrepreneurship’
emphasis versus other emphases within the Business discipline),
and entrepreneurial tendencies (self-identified as having suc-
cessfully started a business). Subject sex (male or female)
and a sex- by-in fect ion int eractio n were also i nclu ded in al l
analyses; for Study 1, we also included a variable for grade
point ave rage (GPA ) to accou nt for gra de-ba sed adm issi on
requirements in different colleges (missing values were r eplaced
with the mean GPA).
(c) Study 3: global patterns of toxoplasmosis
and entrepreneurship
Existing databa ses of T. go n di i infectio n prevalence at the
country level ( the proportio n of samples that teste d positive
for T. go nd ii wi thin a coun try-spec ific survey) were compi led
using publi shed data and revi ews (see elect ronic suppleme n-
tary material ). The data were restricte d to studies condu cted
after 1990. W hen more tha n one T. g ond ii p revalence study
was available for a given c ountry, we used the average ( elec-
tronic sup plementary material , table S2). Because countries
vary in the aver age age at which women get pregnant and thu s
the number of years of potential exposur e to T. gondii , w e used
prevalence estim ates standar dized to an age of 22 years foll owing
Lafferty [13] (also see [28]).
To assess en trepreneurshi p, we used the G EM of entrep re-
neurial activ ity. The GEM databa se is the result of a
worldwide c onsortium o f universitie s (www.gemcons ortium.
org), with data d erived from nationa l surveys of entrepreneu rial
climate, activit ies and attitud es. We used GEM data collec ted
from 2000 to 2010, using the most recent data f or counties
that were surveyed more than once d uring the study p eriod.
Our three depe ndent variab les were the proport ion of peopl e
in a country with en trepreneuria l intention s ( Futsupno ), th e pro-
portion eng aged in entrepren eurial activi ty ( teayy ), and the
proportion in hibited from sta rting a busine ss by their fe ar of
failure ( Frfailop ) (se e electronic su pplementa ry material for
additional variable descriptions). Because all dependent vari-
ables have values that ran ge from zero to one, we us ed
generalized l inear model s with a binom ial distributi on, follow-
ing McDowell & Cox [31]. T he control vari ables used in th e
standard GEM model [32] are d escribed in the electronic sup-
plementar y material, alt hough we ran analyses w ith and
without thes e variables to en sure that findings were robu st
(see electron ic suppleme ntary material ). Overlay ing the GEM
and T. go nd ii data source s yielded a databas e for 42 countr ies
(see electron ic suppleme ntary materia l, table S2). We hypo th-
esized that T. go nd ii prevalen ce would a ssociate po sitively
with the propor tion of peopl e in a given country t hat intend
to start a new busin ess or are c urrently engag ed in ent repre-
neurial activ ities, and ne gatively with the prop ortion of
people inhi bited from startin g a business b y fear of failure.
We did not include int eractions between in fection preva lence
and GEM c ovariates given the lack of ap r i o r i hypotheses and
the desire to avoid o verfitting the models.
3. Results
(a) Study 1: infection patterns in university students
Based on an ana lysis of T. go n di i IgG prevalence from 1495
university stu dents, 22 % tested p ositive. Th ere were 968
confirmed negative cases and 199 samples classified as
‘ambiguous’ even after r etesting, which were r emoved
from furthe r analysis, l eaving 1293 included cases. Using
the 2SD classification method, 195 of 201 known positive
samples (97 % sensitivi ty) and 186 of 1 86 known negative
samples (100% specificity) were classified correctly. Infec-
tion status was a si gnifican t, positi ve predictor o f students’
tendency t o major in bus iness-related f ields ( B ¼ 0.33 ,
s.e. ¼ 0.10, p , 0.005), whi le controllin g for sex and GPA
(althoug h results were compa rable with these c ovariates
removed). Among Bu siness studen ts, 146 of 475 i ndividual s
(31%) tested positive for T. go nd i i IgG, whereas 179 of 818
(22%) of no n-Business maj ors were antibod y-positive. The
odds of being a Business major were 1.4 greater for ant i-
body-positive students [95% CI ¼ (1.1, 1.7)] relative to
those testing negative (figure 1). Female students w ere
also less likely to be Business majors ( B ¼ 2 0.53, s.e. ¼
0.09, p , 0.01), wherea s GPA had no effect . Within the
Business major specifically, students with evidence of
T. gondii exposure were 1.7 more likely to have an estab-
lished emphasis in ‘management and entrepr eneurship’
[95% CI ¼ (1.0, 2.8)] relative to other Busines s subdisciplines
( B ¼ 0.52, s.e. ¼ 0.27, p ¼ 0.054 ( figure 1) . Twenty-four of the
57 (42%) ‘ managemen t and entrepren eurship’ majo rs tested
positive for T. go nd i i compared with 63 o f the 216 (29%) stu-
dents major ing in accounti ng, financ e or marketing. G PA
also associ ated negatively wi th the manag ement/ent repre-
neurship emphasis ( B ¼ 2 0.68, s.e. ¼ 0.29, p , 0.05), while
sex showed no relationshi p. This contrast was even m ore
pronounced in c omparing th e influenc e of infectio n status
on the likeli hood of a concen tration in manage ment/entre -
preneurship ver sus accountin g specific ally ( B ¼ 0.63, s.e. ¼
0.30, p , 0.05). T here were no significa nt interacti ons
between sex and inf ection status for any of th e response
variables, and statisti cs are reported from mod els without
this term.
rspb.royalsocietypublishing.org Proc. R. Soc. B 285 : 20180822
3