Using randomized field experiments to evaluate support programs for older and low-skilled workers with combined survey-register data

在这个项目中，我们使用随机实验来评估劳动力市场计划中两个重要的弱势劳动力群体：老年失业工人以及低技能和老年就业工人。尽管德国劳动力市场在全球金融危机面前表现出弹性，但这两个群体面临的劳动力市场困难在未来可能会加剧。德国的劳动力市场政策包含了许多其他国家不存在的独特项目，因此需要仔细评估。首先，我们调查了关于低技能和老年就业者的公司内部培训补贴的知识在多大程度上改善了他们的劳动力市场结果。其次，我们分析了对老年失业工人的针对性工资支持计划的了解在多大程度上影响了他们的劳动力市场结果。在每种情况下，我们都会检查对就业概率、工资和其他工作特征的影响。此外，我们分析了这些程序产生无谓损失的程度。分析是基于调查和登记相结合的数据。我们的两个实验都是基于信息处理，这是低阈值干预。我们向符合条件的随机选择的治疗组发送信息，告知该组获得支持计划的可能性。其他符合条件的人群，他们没有得到额外的信息，作为对照组。这样的信息处理实验允许两阶段的评估策略。在第一阶段，我们评估了信息提供对平均利益结果的影响。在第二阶段，我们使用信息提供作为工具变量来评估了解项目对兴趣结果的平均影响。两组之间结果变量的差异可能与这些份额的变化有关。具体来说，这允许识别局部平均治疗效果（LATE）和计划参与的平均治疗效果（ATE）的部分识别。一个方法上的挑战是在数据中存在正确审查的失业持续时间的情况下检查对失业后结果的影响。当我们提出一种评估德国劳动力市场计划的创新方法时，我们的结果不仅将为文献和政策结果增加新的见解，而且还将探索未来信息处理和现场实验的潜力和局限性。此外，我们将提供新的理论分析和应用最近发展的基于部分识别的非参数计量经济学方法。

In this project, we use randomized experiments to evaluate labor market programs for two important groups of disadvantaged workers in the labor force: older unemployed workers as well as low-skilled and older employed workers. Despite the resilience of the German labor market in the face of the worldwide financial crisis, these two groups face labor market difficulties that may intensify in the future. German labor market policy contains a number of unique programs that do not exist in other countries and hence call for a careful evaluation. First, we investigate to what extent knowledge about subsidies for in-company training of low-skilled and older employed workers improves their labor market outcomes. Second, we analyze to what extent knowledge about a targeted wage support program for older unemployed workers affects their labor market outcomes. In each case, we examine effects on the employment probability, wages, and other job characteristics. Furthermore, we analyze to what extent these programs produce deadweight losses. The analysis is based on combined survey and register data. Both of our experiments are based on information treatments, which are low-threshold interventions. We send out information to a randomly selected treatment group among the eligible, informing this group about access possibilities to a support program. The other eligibles in the population, who did not receive extra information, serve as the control group. Such information treatment experiments allow for a two-stage evaluation strategy. In the first stage, we assess the effect of the information provision on the average outcomes of interest. In the second stage, we use the information provision as an instrumental variable to assess average effects of knowing about the program on the outcomes of interest. Differences in outcome variables between both groups may then be related to changes in these shares. Specifically, this allows for the identification of local average treatment effects (LATE) and for the partial identification of average treatment effects (ATE) of program participation. A methodological challenge is to examine effects on post-unemployment outcomes in the presence of right-censored unemployment durations in the data.As we propose an innovative approach for the evaluation of German labor market programs, our results will not only add new insights to the literature and policy results, but also explore the potential and limitations of future information treatments and field experiments. Moreover, we will provide novel theoretical analysis and apply recently developed non-parametric econometric methods based on partial identification.