The advice to “fake it ‘til you make it” may, in some contexts, be more useful advice than one might expect (Drago, 2008; Rosenthal & Jacobson, 1968). The present examination of self-report data from 688 recent graduates of 5 colleges (Haverford College, Bryn Mawr College, Santa Clara University, Mills College, and Wellesley College) seeks to test the wisdom of this adage.

3 measures of self-confidence were assessed prior to graduation in the dataset:

• participants’ self-estimated “…ability to succeed in an entry level job relative to other graduating seniors from US colleges and universities” on a 0-100% scale,
• participants’ self-estimated “…ability to compete for promotion and advancement within an employing organization relative to other graduating seniors from US colleges and universities” (also on a 0-100% scale),
• and participants’ self-estimated rank among four fellow study participants in an addition problem task.

Do these measures of self-confidence have positive and significant (at the 5% significance level) effects on annual earnings in participants’ first five years following graduation? Might this differ by gender?

The following OLS regression was estimated to begin to clarify these uncertainties:

• log Earnings_i : a participant’s log-transformed average annual earnings in salary and bonuses in dollars
• log Rank_i : a participant’s log-transformed estimated own rank in the addition task competition
• Promote : a participant’s self-reported confidence in their ability to compete for promotions

This initial model was specified after a prior linear model incorporating confidence in mathematical ability, confidence in ability to earn promotions, and confidence in succeeding at an entry-level position showed no effect of confidence in succeeding at an entry-level position (p > 0.45). The logarithmic transformations of Earnings and Rank were found to be necessary by examining residuals plots and R² values of all possible logarithmic transformations of simple regressions of Earnings on either of the two confidence measures (see below). Robust standard errors were employed to avoid issues arising from heteroscedasticity in the cross-sectional data. The coefficients of interest in this equation are β₁ and β₂ representing the effects of Rank and Promote on Earnings, respectively.

• The estimated value of β₁ (as seen in Table 1), suggests that each additional higher rank participants expected to place in the addition tournament predicted 4.725% in additional annual earnings.
• The estimated value of β₂ suggests that a 10% increase in confidence in being able to compete for promotions is associated with a 5.29% increase in annual earnings.

This equation was then re-estimated separately for female- and male-identifying participants to understand whether the relationship between annual earnings and each of the confidence variables differs by participants’ gender identity (see Table 1).

• The subsequent estimates of β₁ and β₂ suggest that self-confidence in mathematical ability is a significant predictor of annual earnings for women only while confidence in one’s ability to compete for promotions significantly and positively predicts annual earnings for both women and men.
• In particular, women are likely to earn 5.65% more annually for each additional higher rank they expected to place in the addition competition while this effect was not significant for men (p > 0.7).
• Further heterogeneity is evident in the size of the effect of confidence in being promoted on annual earnings. Whereas men are likely to see an additional 8.85% in annual earnings for each increase in confidence in being promoted by 10%, women are likely to earn only 3.11% more annually for each additional 10% boost in confidence in being promoted (this difference was significant, p < .001). 

To understand whether omitted variables may account for the observed effects, however, a second equation was estimated:

W_k : known predictors of annual earnings such as dummy variables indicating

• field of work (business or finance positions, federal government positions, and for-profit work generally)
• role as a supervisor at work
• attainment of a post-graduate degree
• parental status
• identification as heterosexual (to account for anti-LGBTQ pay discrimination)

and continuous variables indicating

• average number of hours the participant works per week
• number of months the participant has spent working
• percentage of childcare that the participant does themself

Notably, as seen in the latter columns of Table 2, the effects of both confidence variables, β₁ and β₂, are not significant when the additional control variables W_k are included (p > .35).

            In summary, OLS regressions of annual earnings on individuals’ confidence in their ability to compete for promotions and their confidence in their ability to compete in addition problem competitions indicated significant positive effects of these confidence measures only when the effects of other predictors of annual earnings were not taken into account. Nonetheless, the estimates from equation 2 (in which the effects of correlates of the self-confidence measures were accounted for but other predictors of earnings were not) suggest that there may be gender heterogeneity in the relationship between self-confidence and annual earnings when that relationship is significant. In particular, women’s confidence in their mathematical ability seems to be more predictive of their earnings than men’s confidence in mathematical ability is for their earnings even as the opposite pattern is evident with respect to individuals’ confidence in being able to compete for promotions. Future investigations might seek to uncover what might explain this gender heterogeneity.