Proponents of school choice argue that if parents were given the right to choose schools, they would choose the more productive ones. Then schools would have to increase productivity or loose students. The result would be an increase in overall school quality. But evidence of the effects of school choice on school productivity is mixed. And students who switch do not gain academically, even if they switch to seemingly better schools (Cullen et al., Cullen/Jacob). One explanation for this lack of evidence is based on information. Official statistics on school quality are in most cases based on pure test outcomes. These output-based measures are mostly driven by the characteristics of the student intake (Wilson). This means that the quality measure of a school reacts strongly to changes in the student population. It also means that such output-based measures reflect the productivity of individual schools poorly. Thus, parents have difficulties to identify the productive schools. An alternative way to measure school quality are value-added measures. These account for differences in student intake and thus reflect school productivity more closely. If value-added measures were used, the mechanism described by proponents of school choice might work (Meyer). Such value-added measures of school quality have been introduced in England and Chicago so far. In England they were largely ignored by parents who had access to dozens of other school quality measures, in Chicago they were introduced only in 2008. Thus, an empirical approach is so far not viable to estimate the effects of value-added quality measures. The drawbacks of output-based measures arise from interactions between the decisions of actors. Theoretical approaches using representative agents are not able to account for such interactions. Thus I follow an approach introduced to the school choice literature by Epple/Romano, Nechyba and Ferreyra. These authors account for interactions by using micro-simulations that represent each household and school as an individual actor. I use such an agent-based computational model to look into the effects of different types of quality measurement on the outcomes of school choice. I focus on the effects that the two information regimes have on (1) gains for switchers, (2) the frequency of school switches, (3) sorting (4) overall school productivity and (5) the correlation between school productivity and quality measures. I add to the literature by modelling the mechanism through which school effort reacts to competition and by using a more sophisticated calibration technique.

I use an agent-based computational model, following Epple/Romano, Nechyba and Ferreyra. In this model households choose a school for their child. This choice is partially based on information about the quality of individual schools. Households are heterogeneous. Innate ability of the children and preferences for school quality vary and are correlated with the socioeconomic background, as found by Hastings et al. and Jacob/Lefgren. Schools choose an effort based on their competition. I calibrate the model with estimations on student-level data from the Chicago Public Schools. Then I simulate the school-choice of several cohorts of students using data of real households and schools. To account for the fact that the parameters are estimated, I run two batches of simulations, one using value-added quality measures, the other using output-based measures. Then I compare the effects of these information regimes by looking at distributions of outcomes generated by the batch runs.

Findings from trial runs of a partially calibrated model: When using output-based measures, switching schools does not lead to academic gains, as parents can hardly identify the more productive schools. There is a high volatility in the system, as quality measures depend on the student intake. If students switch, the quality measures of individual schools change which prompts additional switches. The correlation between school productivity and quality measures is low. School productivity does not react strongly to differences in competition levels, as schools cannot affect their quality levels strongly and thus are not willing to face the costs of increasing effort. When using value-added measures, switching schools does lead to academic gains. Volatility is much lower, as quality measures do not change strongly with switching students. The correlation between quality measures and productivity is higher. School productivity increases more strongly when competition rises. Thus, the overall school quality is higher.