Skip to content

Information for students, faculty and staff regarding COVID-19. (Updated: 31 March 2021)

printicon

Estimation of labor market policy program effects in the presence of interference between individuals

Research project Interference means that the treatment received by one individual may affect the outcomes of other individuals. Although not always explicitly stated, assumptions of no-interference is made routinely by researchers evaluating and estimating effects of treatments, e.g., the effect of job-training on the duration of unemployment.

No-interference is often not a plausible assumption. For example, if a labor market program has an effect on one individual that receives treatments, it will also affect other individuals in the same local labor market, assuming there is a fixed number of jobs. Ignoring interference can lead to misleading conclusions about treatment effects. Thus, it is important to develop methods of estimating treatment (or causal) effects under interference. Our aim is to give recommendations on suitable methods to use in future labor market program evaluations where interference is present.

Head of project

Maria Karlsson
Associate professor
E-mail
Email

Project overview

Project period:

2013-01-01 2014-12-31

Funding

Institutet för arbetsmarknads- och utbildningspolitisk utvärdering (IFAU), 2013-2014: SEK 1,093,000

Participating departments and units at Umeå University

Umeå School of Business, Economics and Statistics

Research area

Statistics

Project description

In the context of causal inference, interference means that the treatment received by one individual may affect the outcomes of other individuals. Although not always explicitly stated, assumptions of no-interference is made routinely by researchers evaluating and estimating effects of treatments, e.g., the effect of job-training on the duration of unemployment. It is often not a plausible assumption. For example, if a labor market program has an effect on one individual that receives treatments, it will also affect other individuals in the same local labor market, assuming there is a fixed number of jobs.

Ignoring interference can lead to misleading conclusions about treatment effects. Thus, it is important to develop methods of estimating causal effects under interference. Also, interference spawns new causal estimands (i.e., causal quantities (effects) of interest), e.g., spillover effects.

Recently the literature on causal inference under interference has begun to grow and in this project the goal is to make a survey of these proposals in order to provide recommendations suitable methods for use in the evaluation of labor market policies. To accomplish this we intend to compare existing proposals by means of simulations of different interference scenarios that are plausible in a labor market. Simulated data will be similar to data available in the Institute for Evaluation of Labour Market and Education Policy (IFAU) registers.

Moreover, we will apply the recommended methods to see if the findings would be different if one takes possible interference into account instead of ignoring it when estimating the effects of a labor market action.

Keywords: causal inference, SUTVA, no-interference, evaluation study, treatment effect