The authors have declared that no competing interests exist.
Conceived and designed the experiments: JN DD GJ MY. Performed the experiments: GA MY. Analyzed the data: JN GA DD GJ. Wrote the paper: JN DD GJ.
Exposure to prenatal androgens affects both future behavior and life choices. However, there is still relatively limited evidence on its effects on academic performance. Moreover, the predicted effect of exposure to prenatal testosterone (T)–which is inversely correlated with the relative length of the second to fourth finger lengths (2D:4D)–would seem to have ambiguous effects on academic achievement since traits like aggressiveness or risk-taking are not uniformly positive for success in school. We provide the first evidence of a non-linear, quadratic, relationship between 2D:4D and academic achievement using samples from Moscow and Manila. We also find that there is a gender differentiated link between various measures of academic achievement and measured digit ratios. These effects are different depending on the field of study, choice of achievement measure, and use of the right hand or left digit ratios. The results seem to be asymmetric between Moscow and Manila where the right (left) hand generates inverted-U (U-shaped) curves in Moscow while the pattern for hands reverses in Manila. Drawing from unusually large and detailed samples of university students in two countries not studied in the digit literature, our work is the first to have a large cross country comparison that includes two groups with very different ethnic compositions.
Performance in schooling is known to be dependent on cognitive ability, family background, and social status, but it is also heavily influenced by biological and psychological traits independent of or even orthogonal to standard notions of cognitive ability. These include aggressiveness or self-confidence, conscientiousness, and/or willingness to take risks. (For cognitive ability, see
Some of these characteristics may derive partly from prenatal exposure to androgenic steroids. The most common marker for measuring prenatal androgens is the second-to-fourth finger digit length ratio (henceforth 2D:4D) with relatively longer fourth fingers (lower 2D:4D) indicating higher fetal androgens
However, the most recent large surveys do not support robust, within-sex correlations between 2D:4D and the masculinity/femininity personality dimensions
Ties to academic achievement are even less well-explored. There is some limited work on the relationship between 2D:4D and academic performance but the findings are mixed and often based on limited samples. Romano found that adult males’ 2D:4D ratios positively predicted examination grades while females’ marks were uncorrelated with these ratios
Brosnan et al.
The strongest claims on 2D:4D effects that might be relevant for understanding academic achievement are to be found in Branas-Garza and Rustichini’s
This work, seen in light of the earlier diverse findings showing at best weak links between 2D:4D and academic performance suggests that there might be strongly sex differentiated effects; further, the unreliable findings across studies could be driven by nonlinearity in the relationship between testosterone and later outcomes. Some characteristics associated with high testosterone could plausibly have non-linear effects on performance – some risk-taking or aggressiveness, for instance, might be beneficial, but too much might lead to destructive behavior (e.g.
Sapienza et al.
The various findings from the literature suggest that the effect of 2D:4D on academic outcomes may be more complex than a linear relationship. There may be many other relevant factors that affect academic outcomes that happen to be correlated with 2D:4D, and any effect of 2D:4D on academic outcomes obtained from a simple linear specification may be an incomplete approximation.
To verify whether the effect could actually be non-linear, and would vary across academic fields and between genders, we specify the following model:
(1)where the academic outcome
To empirically test equation (1), we use two different cross-sectional datasets – one is a sample of over 700 students from the Higher School of Economics (HSE) in Moscow, and the other is a sample of about 120 students from the University of the Philippines School of Economics (UPSE) in Manila. For both Moscow and Manila, all the students in the samples were recruited for the study in a manner consistent with local protocols for human subject research. Though no signed consent forms were obtained, permission for the study was formally obtained at the HSE and the UPSE in accordance with local practice. In addition, the overall survey and research design was reviewed by the George Mason University Office of Research Subjects Protection and it was determined that no review by the Human Subjects Review Board was necessary for participation by the two authors representing GMU who were not directly involved in collecting the survey information presented to them in anonymous form.
In Moscow, measurements of the second and fourth fingers of both the left and right hands of all the subjects were taken by two research assistants using a laser caliper (with the exception of those subjects who had stated in the questionnaire that they had broken their second and/or fourth finger – these were then omitted). In Manila, we had the subjects photocopy their left and right hands, and from these, two research assistants obtained the lengths of the second and fourth fingers using tape measures. Whether by laser caliper or tape measure, finger length is measured as the distance between the middle of the line at the base of the finger up to the point on the fingertip that is perpendicular to that base. Note that in both Moscow and Manila, subjects were allocated among the research assistants, but each assistant measured both the left and right-hand fingers of the subjects assigned to her. Thus, while there may be some variability in the measurements across subjects, we do not expect any biased difference between the measurements of the left hand and the right hand and/or between the measurements of the second and fourth fingers of each hand.
From these measurements, each subject’s digit ratio was computed by dividing the length of the subject’s second finger to the length of her fourth finger, for her left and right hands. For both Moscow and Manila samples, we thus have two proxies for , denoted as Left hand 2D:4D and Right hand 2D:4D.
We use several proxies for individual academic outcomes . For the Moscow sample, we have information on test scores on the college entrance exam–the Unified State Exam (USE)–particularly the Math Score and the Russian (language) Score. (It should be noted that there was an old version of the USE which was in a different form and used a different grading scale. This old USE was taken by the oldest students in the original sample and only as an option, unlike the new version of the USE which is compulsory. To get a consistent set of students for the final sample, we only included the younger students, i.e. those who took the new version of the USE. However, as a robustness check, we also ran regressions using the original sample in which students who took the old USE were included, after re-scaling their scores to approximate the new USE. The results are generally similar to the ones reported in this paper and can be provided upon demand).
We also have data on whether the subject was admitted to HSE based on high scores in pre-college competitions called Olympiads; whether the subject was a recipient of high school honors; and whether the subject was admitted to HSE with a full academic scholarship (virtually all HSE scholarships are based on academic criteria only using non-subjective formulae). For these we constructed the corresponding binary variables Olympiad, High School Honors and Full Scholarship.
For the Manila sample, the subjects provided their grades for all economics courses taken to date and their grades for all mathematics courses taken to date. We converted these to the US grading scale (using the official guidelines of the University of the Philippines) and computed the Economics Weighted Average and the Mathematics Weighted Average according to the University’s convention of using the number of units of the course as its weight.
We also have data on the subjects’ gender from both Moscow and Manila. In addition, because HSE is further divided into different faculties, we create binary variables indicating the particular Faculty to which each Moscow subject belongs: Faculty(Economics), Faculty(Law), Faculty(Management) and Faculty(Political Science).
|
|||||
Variables | N | Mean | Std. Dev. | Min | Max |
Outcome Variables | |||||
Economics Weighted Average (US Score Equivalent) | |||||
Full sample | 121 | 2.542 | 0.672 | 1.250 | 4 |
Females*** | 72 | 2.411 | 0.663 | 1.250 | 4 |
Males*** | 49 | 2.734 | 0.645 | 1.614 | 3.792 |
Mathematics Weighted Average (US Score Equivalent) | |||||
Full sample | 123 | 2.425 | 0.817 | 0.491 | 4 |
Females** | 74 | 2.285 | 0.754 | 0.491 | 4 |
Males** | 49 | 2.635 | 0.871 | 1 | 4 |
Explanatory Variables | |||||
Right hand 2D:4D | |||||
Full sample | 123 | 0.986 | 0.038 | 0.878 | 1.129 |
Females*** | 74 | 0.996 | 0.037 | 0.922 | 1.129 |
Males*** | 49 | 0.971 | 0.034 | 0.878 | 1.043 |
Left hand 2D:4D | |||||
Full sample | 123 | 0.964 | 0.033 | 0.859 | 1.060 |
Females** | 74 | 0.969 | 0.034 | 0.859 | 1.060 |
Males** | 49 | 0.956 | 0.030 | 0.877 | 1.028 |
Female | 123 | 0.602 | 0.492 | 0 | 1 |
Male | 123 | 0.398 | 0.492 | 0 | 1 |
Note: *** 1%, ** 5% Significant difference between the mean values for the female and male subsamples
|
|||||
Variables | N | Mean | Std. Dev. | Min | Max |
Outcome Variables | |||||
Math Score (Unified State Exam) | |||||
Full sample | 277 | 70.484 | 12.292 | 30 | 100 |
Females | 152 | 70.559 | 11.570 | 44 | 95 |
Males | 125 | 70.392 | 13.163 | 30 | 100 |
Russian Score (Unified State Exam) | |||||
Full sample | 421 | 78.948 | 9.287 | 54 | 100 |
Females*** | 242 | 81.277 | 8.714 | 60 | 100 |
Males*** | 179 | 75.799 | 9.129 | 54 | 100 |
Olympiad | |||||
Full sample | 770 | 0.300 | 0.459 | 0 | 1 |
Females | 446 | 0.296 | 0.457 | 0 | 1 |
Males | 323 | 0.307 | 0.462 | 0 | 1 |
High School Honors | |||||
Full sample | 755 | 0.404 | 0.491 | 0 | 1 |
Females*** | 435 | 0.492 | 0.501 | 0 | 1 |
Males*** | 319 | 0.285 | 0.452 | 0 | 1 |
Full Scholarship | |||||
Full sample | 770 | 0.704 | 0.457 | 0 | 1 |
Females | 447 | 0.747 | 0.435 | 0 | 1 |
Males | 322 | 0.643 | 0.480 | 0 | 1 |
Explanatory Variables | |||||
Faculty (Economics) | |||||
Full sample | 796 | 0.273 | 0.446 | 0 | 1 |
Females | 449 | 0.245 | 0.431 | 0 | 1 |
Males | 327 | 0.306 | 0.461 | 0 | 1 |
Faculty (Law) | |||||
Full sample | 796 | 0.319 | 0.466 | 0 | 1 |
Females | 449 | 0.318 | 0.466 | 0 | 1 |
Males | 327 | 0.330 | 0.471 | 0 | 1 |
Faculty (Management) | |||||
Full sample | 796 | 0.205 | 0.404 | 0 | 1 |
Females | 449 | 0.212 | 0.409 | 0 | 1 |
Males | 327 | 0.208 | 0.406 | 0 | 1 |
Faculty (Political Science) | |||||
Full sample | 796 | 0.204 | 0.403 | 0 | 1 |
Females*** | 449 | 0.225 | 0.418 | 0 | 1 |
Males*** | 327 | 0.156 | 0.363 | 0 | 1 |
Right hand 2D:4D | |||||
Full sample | 814 | 0.989 | 0.037 | 0.857 | 1.134 |
Females*** | 449 | 0.994 | 0.038 | 0.857 | 1.134 |
Males*** | 327 | 0.983 | 0.033 | 0.902 | 1.082 |
Left hand 2D:4D | |||||
Full sample | 814 | 0.990 | 0.034 | 0.876 | 1.115 |
Females*** | 449 | 0.995 | 0.035 | 0.890 | 1.115 |
Males*** | 327 | 0.984 | 0.033 | 0.876 | 1.072 |
Female | 776 | 0.579 | 0.494 | 0 | 1 |
Male | 776 | 0.421 | 0.494 | 0 | 1 |
Note: *** 1%, ** 5% Significant difference between the mean values for the female and male subsamples
For both Manila and Moscow, the mean values of the Right hand 2D:4D and Left hand 2D:4D are significantly different for males and females, with females having significantly higher Right hand 2D:4D and Left hand 2D:4D than males. This suggests that, on average, females have significantly less prenatal testosterone exposure than males. In addition, the mean values of Right 2D:4D for females are similar across Manila and Moscow, but the mean Left 2D:4D is lower for females in Manila than in Moscow. Judging only by the left hand, this suggests that female Manila students may have more prenatal testosterone on average than female Moscow students. Male Manila students also may have higher prenatal testosterone than male Moscow students, as the former’s mean Right and Left 2D:4D are lower than the latter’s.
|
|||||
|
Economics Weighted Average (US score equivalent) | Mathematics Weighted Average (US score equivalent) | Right hand 2D:4D | Left hand 2D:4D | Female |
Economics Weighted Average (US score equivalent) | 1.000 | ||||
Mathematics Weighted Average (US score equivalent) | 0.474* | 1.000 | |||
Right hand 2D:4D | 0.069 | 0.055 | 1.000 | ||
Left hand 2D:4D | 0.042 | 0.061 | 0.554* | 1.000 | |
Female | −0.237* | −0.211* | 0.327* | 0.205* | 1.000 |
Note: * Significant at 10%, **5%, ***1%
|
||||||||
|
Economics Weighted Average (US score equivalent) | Mathematics Weighted Average (US score equivalent) | Right hand 2D:4D | Left hand 2D:4D | ||||
Female | Male | Female | Male | Female | Male | Female | Male | |
Economics Weighted Average (US score equivalent) | 1.000 | 1.000 | ||||||
Mathematics Weighted Average (US score equivalent) | 0.429* | 0.472* | 1.000 | 1.000 | ||||
Right hand 2D:4D | 0.149 | 0.182 | 0.215*** | 0.019 | 1.000 | 1.000 | ||
Left hand 2D:4D | 0.059 | 0.145 | 0.048 | 0.202 | 0.559* | 0.466* | 1.000 | 1.000 |
Note: * Significant at 10%, **5%, ***1%
|
||||||||||||
|
Math Score | Russian Score | Olympiad | High School Honors | Full Scholarship | Faculty (Economics) | Faculty (Law) | Faculty (Management) | Faculty (Political Science) | Right hand 2D:4D | Left hand 2D:4D | Female |
Math Score | 1.000 | |||||||||||
Russian Score | 0.404* | 1.000 | ||||||||||
Olympiad | −0.176* | −0.076 | 1.000 | |||||||||
Honors | 0.299* | 0.366* | 0.107* | 1.000 | ||||||||
Full Scholarship | 0.096 | 0.335* | 0.418* | 0.261* | 1.000 | |||||||
Faculty (Economics) | 0.588* | 0.203* | 0.235* | 0.165* | 0.014 | 1.000 | ||||||
Faculty (Law) | −0.396* | −0.103* | −0.099* | −0.033 | 0.091* | −0.419* | 1.000 | |||||
Faculty (Management) | −0.145* | −0.121* | −0.150* | −0.129* | −0.241* | −0.311* | −0.347* | 1.000 | ||||
Faculty (Political Science) | −0.369* | −0.009 | 0.005 | −0.014 | 0.122* | −0.310* | −0.346* | −0.257* | 1.000 | |||
Right hand 2D:4D | 0.005 | 0.022 | 0.015 | 0.053 | 0.070* | −0.054 | 0.009 | 0.117* | −0.068* | 1.000 | ||
Left hand 2D:4D | −0.059 | 0.077 | 0.019 | 0.051 | 0.064* | −0.008 | −0.024 | 0.067* | −0.032 | 0.564* | 1.000 | |
Female | 0.007 | 0.292* | −0.011 | 0.208* | 0.113* | −0.068* | −0.012 | 0.004 | 0.086* | 0.159* | 0.156* | 1.000 |
Note: * Significant at 10%, **5%, ***1%
|
||||||||||||
|
Math Score | Russian Score | Olympiad | High School Honors | Full Scholarship | |||||||
Female | Male | Female | Male | Female | Male | Female | Male | Female | Male | |||
Math Score | 1.000 | 1.000 | ||||||||||
Russian Score | 0.448* | 0.388* | 1.000 | 1.000 | ||||||||
Olympiad | −0.188** | −0.165 | −0.129** | −0.001 | 1.000 | 1.000 | ||||||
Honors | 0.372* | 0.239* | 0.317* | 0.336* | 0.082*** | 0.151* | 1.000 | 1.000 | ||||
Full Scholarship | 0.127 | 0.066 | 0.304* | 0.308* | 0.366* | 0.493* | 0.182* | 0.330* | 1.000 | 1.000 | ||
Faculty (Economics) | 0.623* | 0.553* | 0.252* | 0.191** | 0.199* | 0.281* | 0.219* | 0.142** | −0.014 | 0.066 | ||
Faculty (Law) | −0.410* | −0.383* | −0.064 | −0.131 | −0.077 | −0.126** | −0.053 | 0.005 | 0.079*** | 0.108 | ||
Faculty (Management) | −0.172** | −0.117 | −0.192* | −0.024 | −0.127* | −0.183* | −0.116** | −0.165* | −0.235* | −0.255* | ||
Faculty (Political Science) | −0.362* | −0.382* | −0.037 | −0.064 | 0.005 | 0.007 | −0.051 | −0.005 | 0.154* | 0.057 | ||
Right hand 2D:4D | −0.046 | 0.062 | −0.037 | −0.045 | 0.006 | 0.036 | 0.069 | −0.061 | 0.011 | 0.114** | ||
Left hand 2D:4D | −0.093 | −0.027 | −0.024 | 0.093 | 0.027 | 0.013 | 0.026 | 0.008 | −0.010 | 0.122** | ||
|
|
|
|
|
|
|
||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
Faculty (Economics) | 1.000 | 1.000 | ||||||||||
Faculty (Law) | −0.389* | −0.466* | 1.000 | 1.000 | ||||||||
Faculty (Management) | −0.295* | −0.340* | −0.354* | −0.360* | 1.000 | 1.000 | ||||||
Faculty (Political Science) | −0.307* | −0.285* | −0.368* | −0.302* | −0.279* | −0.220* | 1.000 | 1.000 | ||||
Right hand 2D:4D | −0.033 | −0.085 | −0.029 | 0.081 | 0.147* | 0.071 | −0.079*** | −0.077 | 1.000 | 1.000 | ||
Left hand 2D:4D | 0.024 | −0.041 | −0.067 | 0.049 | 0.115** | −0.007 | −0.063 | −0.004 | 0.574* | 0.542* | 1.000 | 1.000 |
Note: * Significant at 10%, **5%, ***1%
Note also that while Left hand 2D:4D and Right hand 2D:4D are positively correlated in both samples, the correlation is not very high – only 0.55 for Manila and 0.56 for Moscow when aggregating females and males. The correlations appear larger for females than males, with 0.56 for females in Manila and 0.47 for males, and 0.57 for females in Moscow and 0.52 for males. This indicates that prenatal testosterone may be expressed differently between the hands, and between females and males, and suggests that regression results may differ significantly by gender and depending on which hand is used. In fact, Full Scholarship is significantly positively correlated with both Left and Right 2D:4D for males in Moscow, while the Mathematics Weighted Average is correlated with Right 2D:4D for females in Manila.
Gender also appears to have a direct correlation with academic outcomes and digit ratios. In Manila, being female is negatively correlated with the Economics and Mathematics Weighted Average, and positively correlated with Left and Right 2D:4D. Being female is also positively correlated with digit ratios in Moscow, but unlike Manila, it is positively correlated with academic outcome variables, specifically, Russian Score, High School Honors and Full Scholarship.
Lastly, note that the Faculty variables in Moscow are significantly correlated with the academic outcome variables for the full sample, and when subdividing by gender. In particular, Faculty (Economics) is positively related to all outcomes except Full Scholarship, while Faculty (Law), Faculty (Management), and Faculty (Political Science) are negatively related to most of the outcome variables.
The foregoing suggests that gender, choice of Faculty and hand measured can modify the association between digit ratio and academic outcomes.
The Manila graphs depict an inverted-U relationship between Mathematics Weighted Average and Left 2D:4D for females (but not so for males), as well as an inverted-U relationship between Economics Weighted Average and Left 2D:4D for males (but not for females). Note, however, that when the Right 2D:4D is used, the non-linear relationships are now U-shaped for both males and females.
On the other hand, the Moscow graphs generally show inverted-U relationships between various academic outcomes and Right 2D:4D, even within different Faculties (especially for females), while the relationships between outcomes and Left 2D:4D are mostly U-shaped (especially for males).
We now provide more rigorous regression-based tests of our hypothesis that digit ratios affect academic outcomes in a quadratic manner and that the effects are differentiated between genders, Faculty types, and left and right hands.
The next tables report the results from OLS regressions of equation (1) for the continuous dependent variables in Moscow and Manila, i.e. Math Scores, Russian, Scores, Economics Weighted Average and Mathematics Weighted Average. For the binary dependent variables High School Honors, Olympiad and Full Scholarship, equation (1) is interpreted as a Linear Probability Model (LPM). As an alternative to the LPM, logit regressions are also reported. Whenever the OLS/LPM results show that there is a significant quadratic relationship between 2D:4D and the academic outcome, we also compute the optimal value of 2D:4D that maximizes or minimizes this academic outcome. That is, maxima are computed for significant inverted-U relationships, while minima are computed for significant U-shaped relationships. Note that an inverted-U (U-shaped) relationship is implied by a positive (negative) estimated coefficient
|
||||||||
Russian Score | Math Score | |||||||
Female | Male | Female | Male | |||||
Explanatory Variables | OLS | OLS | OLS | OLS | OLS | OLS | OLS | OLS |
Left Hand 2D:4D | −752.148 | −557.914 + | −713.831 | −857.401 | 1455.801 | 641.322 | −1680.256** + | −498.408 |
|
|
|
|
|
|
|
|
|
Square of Left Hand 2D:4D | 374.309 | 257.180 | 374.938 | 453.383 | −744.247 | −335.559 | 849.140** + | 267.912 |
|
|
|
|
|
|
|
|
|
Left Hand 2D:4D × Faculty (Law) | −77.678** + | 20.373 | −33.9398 + | 40.912 | ||||
|
|
|
|
|||||
Left Hand 2D:4D × Faculty (Management) | 16.970 | −57.745 | −21.6169 + | 26.279 | ||||
|
|
|
|
|||||
Left Hand 2D:4D × Faculty (Political Science) | −115.271** + | 45.408 | −49.9604 + | −186.639* + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Law) | 73.579** + | −25.363 | 14.244 | −62.442 + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Management) | −23.475 + | 55.889 | 11.500 | −36.531 + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Political Science) | 112.109** + | −49.721 | 29.306 | 163.886 + | ||||
|
|
|
|
|||||
Constant | 458.627 | 384.719 | 414.762 | 482.390 | −640.098 | −227.823 | 900.529** | 308.294 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
− |
|
|
|
|
|
− |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Lmax (Lmin) is the value of left digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||||
HS Honors | ||||||||||
Female | Male | |||||||||
Explanatory Variables | LPM | Logit | LPM | Logit | LPM | Logit | LPM | Logit | ||
Left Hand 2D:4D | 27.599 | 111.982 | 28.410 | 125.207 | −18.750 | −88.608 | −13.837 | −55.615 | ||
|
|
|
|
|
|
|
|
|||
Square of Left Hand 2D:4D | −13.669 | −55.465 | −14.709 | −64.843 | 9.579 | 45.275 | 6.964 | 27.985 | ||
|
|
|
|
|
|
|
|
|||
Left Hand 2D:4D × Faculty (Law) | −3.206* + | −13.645* + | 0.089 | 0.558 | ||||||
|
|
|
|
|||||||
Left Hand 2D:4D × Faculty (Management) | −0.225 | −0.916 + | −2.282 + | −16.901 + | ||||||
|
|
|
|
|||||||
Left Hand 2D:4D × Faculty (Political Science) | −2.445 + | −10.384 + | −0.366 | −1.741 | ||||||
|
|
|
|
|||||||
Square of Left Hand 2D:4D × Faculty (Law) | 2.993* + | 12.748* + | −0.183 | −0.983 | ||||||
|
|
|
|
|||||||
Square of Left Hand 2D:4D × Faculty (Management) | −0.073 + | −0.329 | 2.072 | 15.728 | ||||||
|
|
|
|
|||||||
Square of Left Hand 2D:4D × Faculty (Political Science) | 2.213 | 9.413 | 0.274 | 1.322 | ||||||
|
|
|
|
|||||||
Constant | −13.420 | −56.475 | −13.002 | −59.522 | 9.449 | 42.382 | 7.243 | 27.100 | ||
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
||||||
|
− |
|
− |
|
||||||
|
|
|
|
|
||||||
|
|
|
|
|
||||||
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
||||||
|
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Lmax (Lmin) is the value of left digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||
Olympiad | ||||||||
Female | Male | |||||||
Explanatory Variables | LPM | Logit | LPM | Logit | LPM | Logit | LPM | Logit |
Left Hand 2D:4D | −4.014 | −17.866 | −7.767 | −5.199 | −63.072* + | −276.595** + | −50.317* + | −239.918* + |
|
|
|
|
|
|
|
|
|
Square of Left Hand 2D:4D | 2.194 | 9.818 | 4.536 + | 5.223 | 32.128* + | 140.909** + | 25.927* + | 123.339* + |
|
|
|
|
|
|
|
|
|
Left Hand 2D:4D × Faculty (Law) | −1.020 + | −6.764 + | 0.639 | 2.746 | ||||
|
|
|
|
|||||
Left Hand 2D:4D × Faculty (Management) | 2.478 + | 14.153 | 0.077 | −0.209 | ||||
|
|
|
|
|||||
Left Hand 2D:4D × Faculty (Political Science) | 2.420 + | 11.038 | −2.712 + | −11.970 + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Law) | 0.819 | 5.844 | −0.930 + | −4.067 + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Management) | −2.747 + | −15.486 + | −0.446 + | −1.665 + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Political Science) | −2.591 + | −11.789 + | 2.582 | 11.426 | ||||
|
|
|
|
|||||
Constant | 2.115 | 7.178 | 3.685 | −0.193 | 31.227* | 134.752** | 24.875* | 116.502* |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
− |
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Lmax (Lmin) is the value of left digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||
Full Scholarship | ||||||||
Female | Male | |||||||
Explanatory Variables | LPM | Logit | LPM | Logit | LPM | Logit | LPM | Logit |
Left Hand 2D:4D | 2.126 | 10.931 | 0.184 | −11.150 | −49.158* + | −259.535* + | −46.485* + | −277.004* + |
|
|
|
|
|
|
|
|
|
Square of Left Hand 2D:4D | −1.130 | −5.816 | 0.279 | 7.531 | 25.859* + | 136.198* + | 24.594* + | 146.418* + |
|
|
|
|
|
|
|
|
|
Left Hand 2D:4D × Faculty (Law) | −0.072 | −1.482 | 0.457 | 2.986 | ||||
|
|
|
|
|||||
Left Hand 2D:4D × Faculty (Management) | 1.885 | 8.420 | 0.189 | 3.741 | ||||
|
|
|
|
|||||
Left Hand 2D:4D × Faculty (Political Science) | 0.699 + | 2.775 + | −0.578 | −4.404 | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Law) | 0.138 | 1.868 | −0.442 | −2.959 | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Management) | −2.070 + | −9.243 + | −0.488 + | −5.083 + | ||||
|
|
|
|
|||||
Square of Left Hand 2D:4D × Faculty (Political Science) | −0.564 | −1.875 | 0.620 | 4.642 | ||||
|
|
|
|
|||||
Constant | −0.248 | −4.028 | 0.276 | 4.657 | 23.948* | 123.967* | 22.590* | 131.488* |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
− |
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Lmax (Lmin) is the value of left digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
Note that the results from logit regressions are similar to the LPM – dividing the logit estimates by 4 gives values that are close to the LPM estimates. The backward stepwise regressions generally confirm the results, as variables are significant whenever they are significant in the OLS/LPM/logit regressions. However, the backward stepwise regressions seem to yield more significant results - see, for instance, female and male Math Scores, male HS Honors, female Olympiads, and female Full Scholarship. This indicates that the variables are
It can be seen that the significant relationships are U-shaped and mostly hold for male students. Without controlling for Faculty type, there are significant U-shaped relationships between Left 2D:4D and Math Score, Olympiad and Full Scholarship for the male subsample. When we further break down the sample by Faculty type, we find significant U-shaped relationships between Left 2D:4D and the following: High School Honors for female law students; Olympiad for male economics students; Russian Score for female law and female political science students; and Full Scholarship for male economics students. Note, however, that for Russian Scores for female law and female political science students, the computed Lmin values lie outside the range of Left 2D:4D values in the Moscow sample. For law, Lmin is below the lowest 2D:4D, which indicates that the sample is on the upward-sloping part of the U-curve; while for political science, the sample is on the downward-sloping part of the U-curve (since their Lmin is above the highest 2D:4D in the sample).
|
||||||||
Russian Score | Math Score | |||||||
Female | Male | Female | Male | |||||
Explanatory Variables | OLS | OLS | OLS | OLS | OLS | OLS | OLS | OLS |
Right Hand 2D:4D | 1219.507*** + | 1149.561* + | −358.210 | −594.132 | 1462.162** + | 1548.033*** + | −833.586 | −386.7195 |
|
|
|
|
|
|
|
|
|
Square of Right Hand 2D:4D | −614.585*** + | −595.052* + | 176.008 | 304.791 | −738.570** + | −771.270*** + | 436.514 | 232.749 + |
|
|
|
|
|
|
|
|
|
Right Hand 2D:4D × Faculty (Law) | −57.484 + | 11.199 | 3.351 | −89.117 + | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Management) | −52.595 + | 24.576 | −0.222 | 54.965 | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Political Science) | −83.038** + | −46.469 | 51.339 | −27.447 + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Law) | 53.448 | −15.820 + | −23.801 + | 67.591 | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Management) | 46.244 | −27.884 | −9.7882 + | −66.265 + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Political Science) | 79.941** + | 44.229 | −73.334 + | 2.309 | ||||
|
|
|
|
|||||
Constant | −522.899 | −469.594** | 257.593 | 367.324 | −652.064*** | −698.040** | 467.597 | 233.046 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
− |
|
|
|
− |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Rmax (Rmin) is the value of right digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||
HS Honors | ||||||||
Female | Male | |||||||
Explanatory Variables | LPM | Logit | LPM | Logit | LPM | Logit | LPM | Logit |
Right Hand 2D:4D | 28.398* + | 122.204 | 25.317* + | 123.042 | −1.684 | −1.054 | −20.747 | −70.781 |
|
|
|
|
|
|
|
|
|
Square of Right Hand 2D:4D | −13.739* + | −59.180 | −12.610* + | −61.225 | 0.434 | −1.549 | 10.263 | 34.745 |
|
|
|
|
|
|
|
|
|
Right Hand 2D:4D × Faculty (Law) | −3.279** + | −14.069* + | 1.805 + | 9.223 | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Management) | −0.928 + | −4.080 + | −0.726 + | −3.227 + | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Political Science) | −1.053 + | −4.249 + | −5.792** + | −28.893** + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Law) | 3.061* + | 13.146* + | −1.918 + | −9.778 | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Management) | 0.623 | 2.804 | 0.488 | 1.899 | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Political Science) | 0.818 | 3.264 | 5.820** + | 28.989** + | ||||
|
|
|
|
|||||
Constant | −14.141*** | −62.948 | −12.003 | −60.940 | 1.520 | 1.611 | 2405398.000 | 35.468 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
− |
|
||||
|
|
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
||||||
|
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Rmax (Rmin) is the value of right digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||
Olympiad | ||||||||
Female | Male | |||||||
Explanatory Variables | LPM | Logit | LPM | Logit | LPM | Logit | LPM | Logit |
Right Hand 2D:4D | −0.175 | −0.7937 | −0.955 | 31.4653 | 50.850* + | 266.6381*** + | 16.612 + | 157.5927 + |
|
|
|
|
|
|
|
|
|
Square of Right Hand 2D:4D | 0.122 | 0.5637 | 0.640 | −14.9175 | −25.542* + | −133.9596*** + | −7.014 | −74.4144 |
|
(41.0894) |
|
(43.1791) |
|
(77.4587) |
|
|
|
Right Hand 2D:4D × Faculty (Law) | −2.073 + | −11.1257 + | 3.421* + | 16.2373*** + | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Management) | −0.790 + | −5.1301 + | 0.916 + | −3.1928 + | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Political Science) | 2.576 + | 14.5085 | −1.593 + | −10.1153 + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Law) | 1.872 + | 10.1934 + | −3.768* + | −17.8090*** + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Management) | 0.512 | 3.7685 | −1.318 | 1.2536 | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Political Science) | −2.754 + | −15.3794 + | 1.438 | 9.4827 | ||||
|
|
|
|
|||||
Constant | 0.349 | −0.6357 | 0.771 | −16.6962 | −24.969*** | −133.3367*** | −9.031 | −82.8869 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
− |
|
|
|
||||
|
|
|
|
|
||||
|
||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
||||||
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Rmax (Rmin) is the value of right digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||
Full Scholarship | ||||||||
Female | Male | |||||||
Explanatory Variables | LPM | Logit | LPM | Logit | LPM | Logit | LPM | Logit |
Right Hand 2D:4D | 30.968* + | 145.569* + | 27.158* | 147.098* | −13.4755 | −82.8777 | −40.81 | −250.6419 |
|
|
|
|
|
|
|
|
|
Square of Right Hand 2D:4D | −15.412* + | −72.427* + | −13.358 | −72.359*** | 7.666 + | 45.855 + | 22.528 + | 137.110 + |
|
|
|
|
|
|
|
|
|
Right Hand 2D:4D × Faculty (Law) | 0.279 | 1.397 | 3.523* + | 19.050* + | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Management) | −1.835 + | −7.719 + | 1.157 | 9.603 | ||||
|
|
|
|
|||||
Right Hand 2D:4D × Faculty (Political Science) | 1.118 + | 6.961 + | −0.421 | −3.889 | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Law) | −0.219 | −1.056 | −3.576* + | −19.414* + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Management) | 1.635 | 6.824 | −1.505 + | −11.232 + | ||||
|
|
|
|
|||||
Square of Right Hand 2D:4D × Faculty (Political Science) | −0.990 | −6.109 | 0.459 | 4.141 | ||||
|
|
|
|
|||||
Constant | −14.785 | −71.939*** | −13.041 | −73.586*** | 6.474 | 37.712 | 19.032 | 114.712 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|||||||
|
|
Note: The numbers in brackets are robust standard errors; * Significant at 10%, ** 5%, *** 1% in OLS, LPM or Logit regressions; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Rmax (Rmin) is the value of right digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values in OLS and LPM regressions.
|
||||||||
Economics Weighted Average | Mathematics Weighted Average | |||||||
Female | Male | Female | Male | |||||
Explanatory Variables | OLS | OLS | OLS | OLS | ||||
Left hand 2D:4D | 1.175 | 5.067 | 3.127 | 141.498 | 1.076 | 106.71 *** | 5.865 | 25.084 |
|
|
|
|
|
|
|
|
|
Square of Left hand 2D:4D | −2.015 | −72.476 | −54.615*** | −10.066 | ||||
|
|
|
|
|||||
Constant | 1.273 | −0.604 | −0.254 | −72.476 | 1.242 | −49.774 | −2.970 | −12.134 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
− |
− |
|
|
− |
− |
|
− |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
0.977 |
Note: The numbers in brackets are OLS standard errors; *Significant at 10%, ** 5%, *** 1%; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Lmax (Lmin) is the value of left digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values.
|
||||||||
Economics Weighted Average | Mathematics Weighted Average | |||||||
Female | Male | Female | Male | |||||
Explanatory Variables | OLS | OLS | OLS | OLS | ||||
Right hand 2D:4D | 2.614 | −113.175*** + | 3.450 | −177.03* + | 4.3406* | −94.946*** | 0.484 | −197.373*** |
|
|
|
|
|
|
|
|
|
Square of Right hand 2D:4D | 57.595*** + | 94.147* + | 49.386*** + | 103.215*** | ||||
|
|
|
|
|||||
Constant | −0.193 | 57.595 | −0.614 | 85.755 | −2.038 | 47.793 | 2.166 | 96.854 |
|
|
|
|
|
|
|
|
|
|
0.022 | 0.063 |
|
0.0996 | 0.046 | 0.0685 |
|
0.0442 |
|
|
|
|
|
|
|
− |
|
|
|
|
|
|
|
|
|
|
|
72 | 72 |
|
|
|
74 |
|
49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Note: The numbers in brackets are OLS standard errors; * Significant at 10%, ** 5%, *** 1%; + p-value equal or less than 0.10 tolerance level in backward stepwise regressions, implying that the variable ought not to be removed from the model. Rmax (Rmin) is the value of right digit ratio that maximizes (minimizes) the dependent variable, equal to -b1/(2×b2), computed only for significant values.
All these results indicate the following patterns across Manila and Moscow. In Moscow, using the right (left) hand generates inverted-U (U-shaped) curves while in Manila, using the left (right) hand generates the inverted-U (U-shape).That is, without accounting for gender, the results for Manila are opposite of those for Moscow depending on which hand is used. However, when we consider gender subsamples, both Manila and Moscow seem to produce a consistent trend in that the U-shaped curve seems to be more associated with male students. In Manila, while Right 2D:4D also generates U-shaped curves for females, note that the only significant results for males are U-shaped. In Moscow, it seems that irrespective of which hand is used, the significant results for males are almost always U-shaped.
We have shown in both Moscow and Manila that the degree to which prenatal testosterone is linked to academic achievement exhibits some nonlinearity, and the precise relationship is dependent on gender, faculty, or subject choice, and on which hand is used to proxy for prenatal testosterone.
To the extent we do not yet understand the precise mechanism through which prenatal androgens manifest themselves in the right versus the left hand, this suggests that much more needs to be done to learn how we can use these measures to study the effects of prenatal testosterone on achievement. Our research combined with the findings of
We would like to thank Karen Lazaro, Ekaterina Orel, DilyaraValeeva, Ekaterina Kochergina, and Filipinas Bundoc for excellent research assistance. We are grateful to Aldo Rustichini and to two anonymous referees for comments and suggestions.