SAT-like financialproficiency test.

While we do not have access to individual student-levelor class-level data, we were provided aggregated information at the grade level.

Wefind some evidence for being behind on repayments in follow-up 1, however, thecoefficient is not statistically significant with our full set of controls in panel C.

The results show that students in treated schools had a significantly higher measuredintention to save (51) than those in control group (49) at follow-up 1, and 53 com-pared to 51 in follow-up 2. Normalizing by the standard deviations in the controlgroup, these effect sizes represent 9–12 percent of a standard deviation.

the available sources include banks or financial institutions which, typicallyissue credit cards, retail stores, which typically sell items on installments, and fam-ily or friends, which typically lend money informally.

2 per-centage point greater likelihood of treated students working as employees in bothfollow-up surveys (columns 3 and 4). Additionally, columns 5 and 6 show a 4–5 per-centage point higher likelihood of earning an income among treated students.

First, we consider the hypothesis that making students focus on savings and bud-geting made them more aware of money, which in turn led them to spend more.

we consider whether students faced a multi-tasking problem in thatthe financial education curriculum urged them to save, budget, and spend wiselyall at the same time

Thepercentage of parents who save more than 0 increased from 76 percent in controlschools to 78 percent in treatment schools. The average percentage of income savedincreased from 12 percent in control schools to close to 13 percent in treatmentschools. Parents in student treatment schools were also more likely to list monthlyexpenses in a budget, with an increase from 37 percent of parents in the controlschools to 39 percent of parents in the treatment schools.

We can directly test whether parents who attended the financial educationworkshop had influence over the decisions and financial choices of their children,compared to parents who attended the health education workshops

simultaneous surveys in all schools in our sample.

Comparing the lists of current students we received from schools before theworkshop to the filled-in exit questionnaires gives an average attendance rate of46 percent across the 109 schools that returned questionnaires.

We received1,553 filled-in parent questionnaires from 109 treatment schools, implying thatparent workshops did not take place in the other 153 treatment schools that hadoriginally provided a current list of their students for the parent intervention (theseschools either did not organize the workshop or parents did not attend the work-shop).

Six suchmeetings were held in September and October 2011, with teachers and educatorsfrom all six states. All meetings were conducted in facilities provided by the stateeducation departments and meals were provided. The meetings were generally wellattended (more than 200 educators attended in São Paulo), and lasted an average offour hours each

Over 95 percent of treatment school principals report that they received the text-books for the first semester, and 93 percent report receiving them for semesters 2and/or 3.

Panel C additionally controls for baseline values of the dependent variable,as per McKenzie (2012), and for student gender (see Section VIC). When baselinevalues have missing observations, we replace these with zero and include a dummyvariable indicating that the observation was missing.

All analysis tables have a standard reporting format where panel A reportscross-sectional regression results from estimating equation 1 without control vari-ables. Panel B controls for school pair dummies to account for our method ofrandomization, as per Bruhn and McKenzie (2009). Following the suggestion inGlennerster and Takavarasha (2013), we include a dummy for each school pair werandomized, unless doing so would lead us to drop the pair (due to noncompliance ofsome schools or re-randomization as described in Section IVB). We instead includea common dummy for all schools that would have been dropped with a school pairdummy.

Only schools that sent the list were included in the study as thiswas taken as a signal that they were willing to implement the workshop. A total of264 treatment schools provided a list, covering 8,534 students.

44 percent

Brazil is divided into 26 states and a Federal District. The Federal District andfive states were part of the study, including three of the most populous and devel-oped states (São Paulo, Minas Gerais, and Rio de Janeiro), and two states (Cearáand Tocantins) that represent less developed areas of Brazil.

To

Of particular interest iswhether a gender difference in tournament entry is explained bygeneral factors such as overconfidence, risk, and feedback aver-sion (Explanations 2– 4) or if part of such a difference is accountedfor by preference differences for performing in a competition(Explanation 1).

What distinguishes Explanation 1 from the otherthree is that it relies critically on the tournament-entry decisionresulting in a subsequent competitive performance

or example, Lunde-berg, Fox, and Puncochar [1994] argue that the reason why somestudies do not find gender differences in confidence on generalknowledge is because it is not in the masculine domain

In addition to suggesting that men hold a stronger preferencefor competition these evolutionary explanations are also used toexplain why men often are more confident in their relative per-formance and less averse to risk.

One argues that since men canhave many more children than women the potential gain inreproductive success from winning a competition is much greaterfor men, and men have therefore evolved to be more competitivethan women [Daly and Wilson 1983]. The second theory focuseson one gender being responsible for parental care. While a man’sdeath does not influence his current reproductive success, a wom-an’s death may cause the loss of her current offspring [Campbell2002]. Thus differences in potential losses as well as potentialgains from competition may make males more eager to compete.

Nurture as well as nature may cause women to be relativelymore reluctant to perform in a competition. First, we tend to raisegirls and boys differently

Prior research suggeststhat performance measures are particularly important in thisenvironment as men and women who perform similarly in non-competitive environments can differ in their performance whenthey have to compete against one another (see Gneezy, Niederle,and Rustichini [2003], Gneezy and Rustichini [2004], and Larson[2005])

Finally, controlling for gender differences in general factorssuch as overconfidence, risk, and feedback aversion, we estimatethe size of the residual gender differenc

also determine if, absent the thrill orfear of performing in a competition, a gender gap in choice ofcompensation scheme still occurs

Compared to payoff-max-imizing choices, low-ability men enter the tournament too much,and high-ability women do not enter it enough

factors suchas risk and feedback aversion only play a negligible role

solve a real task