Former graduate students

Boys will always be boys? The impacts of gender-based Affirmative Action and Role Models on competitiveness in the lab

Gender differences in labour market outcomes are pervasive in current society. Competitiveness is viewed as a possible factor contributing to the favourable labour market outcomes for men, with the stylised fact being that men are more competitive than women on average. In this study, we experimentally investigate whether institutional structures (gender-based Affirmative Action and Role Models) mitigate the gender differences in competitiveness. Although Affirmative Action has been explored in the literature, the longer-term implications of the preferential treatment creating role model representation have not been examined. Consistent with the literature, we find men have a higher preference for a competitive environment compared with women in our sample. However, there are no significant impacts of the gender based institutional structures on the choice to compete in our experiment. When analysing the responsiveness of performance to a competitive environment and the associated beliefs, we see the female role model treatment encourages a competitive spirit in everyone, whereas the Affirmative Action treatment has a negative effect on the competitive performance of African men. In this experiment, gender-based institutions therefore either have unexpected effects of encouraging competitiveness in all participants, or inadvertently reinforce competitiveness gaps across other identities dimensions, such as population-group. One therefore needs to be considerate of other dimensions of identity in addition to gender when devising preferential treatment policies, and the resulting role model representation, in practice. Changes in beliefs can only partially be exercised as an explanatory channel for these effects.

Words can change worlds: An impact evaluation of Shine Literacy

 The recent PIRLS and SACMEQ reports painted a dire picture of South Africa’s literacy situation: 80% of Grade 4 learners were rendered unable to read for meaning and 27% of Grade 6 learners as functionally illiterate. These results need to be contextualized against the extensive public spending that is incurred on education. Though it appears that learners are in school, they do not seem to be learning, a phenomenon encountered repeatedly in the developing world. The production process of educational outputs is often being hidden in the black box.

With a large body of research confirming how reading literacy holds predictive validity for later child development and academic success, Shine Literacy offers an intervention that is set at lower primary school level and is intended as a swift corrective measure for those who struggle to read early on. This dissertation conducted a quasi-experimental impact evaluation by estimating the treatment effect of Shine Literacy via difference-in-differences and propensity score matching. By using the available data which included Shine’s diagnostic test scores, attendance data and Grade 3 Systemic test score data obtained from the Western Cape Department of Education, the estimation procedures arrived at average treatment effects ranging between 0.6 to 1.9 standard deviations. IsiXhosa and “At risk” learners capture the largest test score improvements, and therefore are the main beneficiaries of the programme. This renders Shine Literacy as an extremely valuable input in the production of better literacy and thereby better schooling outcomes. It helps those at the bottom end of the distribution. Furthermore, positive impact on Systemic Mathematics test scores was found as well, confirming the predictive power of literacy on numeracy repeatedly discussed in the literature.

A low-cost, low-intensity contingency management smoking cessation programme with students: Experimental evidence

Tobacco consumption is a pressing global issue, leading to more than five million deaths each year. In South Africa, the smoking prevalence rate is stubbornly high, implying that a successful smoking cessation programme could have large social benefits, particularly if it targets young smokers. Contingency management interventions, which provide cash transfers conditional on biochemically-verified abstinence, have been effective in bringing about increased smoking cessation rates. However, contingency management programmes are typically very costly and involve frequent monitoring. This dissertation presents results of randomised controlled trial evaluating a low-cost, low-intensity contingency management smoking cessation programme conducted on a sample of treatment-seeking student smokers at the University of Cape Town in 2017. 

There is a statistically significant treatment effect, that is robust across multiple specifications, which increases the likelihood of abstinence by 13- 20%. In addition, the programme as a whole decreased the smoking intensity of non-abstainers. This study suggests, therefore, that a low-cost, low-intensity contingency management smoking cessation programme is efficacious in promoting abstinence amongst treatment- seeking students, and that it should be added to the tobacco control toolkit in South Africa.

The notion of trust has great significance to an economy. Trust is known to be associated with efficient judicial systems, improved government functioning with lower corruption, and better financial outcomes (Johnson and Mislin, 2011). However, many researchers have argued that risk attitudes may confound the measurement of trust because trusting decisions involve outcomes that have only some probability of occurring.

This study therefore seeks to question whether risk attitudes predict trusting decisions in the Berg, Dickhaut and McCabe (1995) Investment Game amongst students at the University of Cape Town in 2016. The statistical method adopted is maximum likelihood estimation which accounts for subject errors in decision making.

This study finds that having additional information on the past behaviour of trustees does not affect the trusting behaviour of trustors. In addition, the presence of a human trustee, versus a computer, is found to significantly influence behaviour and decisions made by trustors in the trust game.

It is also found that subjects are, on average, risk averse with 62% of subjects exhibiting high levels of risk aversion, and females being more risk averse than males. Subjects were also found to subjectively distort probabilities, where subjects would overweight low probabilities and underweight moderate to high probabilities.

Expected Utility models and Rank-Dependent Utility models show that risk and trust are statistically significantly related and that the reasons for trusting one's partner may have arisen out of an inner need to simply trust that person. In addition, risk preferences were able to predict trusting decisions in the environment of risk and the environment of trust. Risk and trust therefore go hand-in-hand and it can be argued that trusting decisions are perceived as decisions involving risk. This study therefore finds that trusting decisions are in fact confounded by risk attitudes, so that a subject may be seen as trusting when actually they are just risk-seeking, or seen as non-trusting when they are just simply risk averse.

Getting the message: Using parental text messaging to increase learner attendance

This paper presents results from a randomised controlled trial in low-income neighbourhoods in Cape Town, South Africa, to test whether parental messages can increase learner attendance at after-school programmes. Parents that were randomly assigned to one of two treatment groups received simple weekly text messages providing them with information about their children’s attendance in the previous week. Learners whose parents received text messages attended 5.6%-6.1% more after-school sessions than the control group (p<0.01) on average, after controlling for background characteristics and spillover effects. This effect was sustained over the course of the observation period.

Structured interviews with parents suggest that those parents who received messages were more likely to engage their children regarding the after-school programme, and were better able to monitor their children’s attendance. The intervention cost approximately 0.07 USD per child per week. This paper shows that low-cost text messages to poorer parents can increase learners’ investment in their education, and shows potential to be scaled up.

Developing consumption feedback principles for monthly utility bills informed by behavioural economics: Evidence from controlled experiments

•  Better feedback principles for the utility bills in South Africa need to be developed. Utility providers might be able to "nudge" consumers towards more desirable consumption patterns by delivering simpler and better feedback informed by applied behavioural sciences. Two sets of controlled experiments were conducted with over 1,500 subjects to identify bill design strategies that could overcome two major barriers to effective consumption feedback:
• The complexity of the utility bill, especially with regards to tariff calculations, and
• Consumer’s declining mindfulness of utility consumption between billing moments.

The “Utility Bill Redesign” experiment, using a randomised control trial, investigates how improving billing feedback design increases consumer’s understanding of energy usage and costs. More than 1,300 participants are randomly assigned to different treatment groups and receive one of nine redesigned utility bills or the current standard bill. Thereafter, participant’s understanding of the bill they received is tested through a questionnaire. We find that restructuring the bill in a logical order and displaying the amount of electricity consumed in each tariff block with separate bar graphs is a successful way to increase consumer understanding of the bill, especially with regards to the step tariff. Further, the results clearly show that consumers are unable to make sense of a utility bill that is not in their home language, even when adding utility specific symbols.

We conclude that significant low-cost improvements can be made to utility bills to increase consumer comprehension.

In the “Attention Redirection” experiment, participants are assigned to different treatment groups and are given an online task that requires daily attention and effort in order to maximise pay-offs. We find that daily SMS reminders significantly redirect attention to the daily task. A blank graph, given to participants at the beginning of the experiment to assist them in self-managing their behaviour, has no effect on task adherence. The results illustrate how inattention routinely leads to sub-optimal behaviour in a specific task area and the resulting welfare loss. A purely bill-based strategy is rendered unsuccessful.

Does social exclusion and power asymmetries reduce cooperation?

 In many societies around the world, power asymmetries exist between individuals as a result of discriminatory institutions, which in some cases give rise to significant wealth inequality. These discriminatory origins of inequality may result in strained social relationships between the powerful and the powerless. Moreover, even after the inclusion of individuals into some aspects of society, the legacy and effects of past exclusion can be difficult to reverse. 

This thesis studies these effects by looking at the relationship between social exclusion, in a political and economic context, and other-regarding preferences. In particular, it analyses the impact of historically exploitative origins of inequality. This is done through a novel experiment which uses two versions of the Prisoner's Dilemma game, one with power asymmetries and the other without, to engender inequality in a Public Goods game. The experiment affords political power to some players and not to others, thereby creating an exclusionary setting which can lead to inequality depending on how individuals choose to use their power. The design aims to isolate the effects both of the proclivity to use power and the manner in which power is used.

The results show that power asymmetries are more often used to ensure equal outcomes in the PD game than to advantage oneself at the expense of another player. Compared to randomly allocated unequal endowments in the Public Goods game, a history of prior play reduces contributions to the public account. There are no significant differences found between treatments in which the opportunity to exploit is exogenously or endogenously determined.

Disaggregating the results shows that rich players free-ride off the poorer players when prior player is characterised by equal power between players. Moreover, players who were without power in prior player and were matched with a co-player who chose to exploit their disadvantage for monetary gain, contribute less to the public good. This effect is conditional on the powerful player benefiting, at the expense of the powerless, monetarily from the asymmetry in power.

This study concludes that the origins of inequality can have important effects on cooperation and other-regarding preferences in socially exclusive contexts. Moreover, they indicate that the origins of inequality, derived from previous interactions, have significant effects even when subsequent interactions are with anonymous strangers. This suggests that addressing inequality is important for social cohesion and welfare in contexts of past injustices. To the best of the author's knowledge, this is the first study to explore the discriminatory origins of inequality and social preferences adding a new perspective to the study of social integration in (previously) unjust societies.

Stochastic Models in Experimental Economics

Shortly after the introduction of Expected Utility Theory (EUT), economists and psychologists began publishing results that showed choices made by experimental subjects which apparently violate one or more of the EUT axioms. I discuss economists’ responses to this evidence. These vary from developing new theoretical models, models that nest EUT as a special case, such as Rank Dependent Utility (RDU) and Regret Theory, as well as models that do not nest EUT, such as Cumulative Prospect Theory, to critiques of experimental methods and scope, to the promotion of stochastic models of choice. I discuss popular stochastic choice models in depth and evaluate their normative coherence. I find that the “Random Preferences” stochastic model fails to make normatively coherent statements, in contrast to the “Random Error” and “Tremble” models, which do so.

I demonstrate a method to calculate the unconditional likelihood of choice errors for populations of EUT-compliant and RDU-compliant agents, and show how certain characteristics of the population relate to the likelihood of these choice errors and their costliness in terms of forgone welfare. I find that elements of the stochastic model that are not related to preference relations tend to have a greater influence on unconditional welfare estimates than the preference parameters themselves. Finally, I conduct a power analysis of the ability of a lottery battery instrument to correctly classify experimental subjects as employing either EUT or RDU, and the effect of this classification on the accuracy of the estimates of welfare surplus for the subjects. 

For large ranges of parameter values for these models, I find that the probability of type I and type II errors in the classification process are not trivial, and can be very costly in terms of welfare surplus. Additionally, I show that for a hypothetical population comprising subjects employing EUT or RDU, we can arrive at more accurate welfare surplus estimates in aggregate by assuming that every subject employs the RDU functional, rather than by first trying to differentiate RDU subjects from EUT subjects.