Everything is Endogenous But…

Design Based Approach

Causal inference is an important concept in many fields, particularly in the social sciences and medical research. The design-based approach to causal inference is one of the most widely used methods to investigate causal relationships between variables.

This approach is based on manipulating a treatment or exposure of interest and then comparing the outcomes between a treatment group and a control group. Randomized controlled trials are an example of this approach, where participants are randomly assigned to either the treatment or control group. By randomly assigning participants, the assumption is that there are no systematic differences between the treatment and control groups, and any observed differences in outcomes can be attributed to the treatment or exposure of interest.

However, the design-based approach has limitations. For example, it may not be feasible or ethical to manipulate certain variables of interest in a study. In addition, the design-based approach assumes that the treatment or exposure of interest is the only factor affecting the outcome variable, which may not be the case in real-world settings.

Despite these limitations, the design-based approach remains a powerful tool for investigating causal relationships. It provides a useful framework for designing experiments and interpreting the results of randomized controlled trials. By carefully manipulating the treatment or exposure of interest and comparing outcomes between treatment and control groups, researchers can gain valuable insights into the causal relationships between variables of interest.

Taking the Selection Bias Out: Instrumental Variable

In addition to the design-based approach, another way to handle selection bias in observational studies is to use instrumental variables. This approach is based on the idea that there may be a variable that is related to the exposure or treatment of interest, but is not directly related to the outcome of interest. By using an instrumental variable, researchers can estimate the causal effect of the exposure or treatment on the outcome variable, while controlling for confounding variables that may be associated with both the exposure and the outcome.

However, identifying a valid instrumental variable can be challenging, and the technique requires a deep understanding of the underlying causal mechanisms for the exposure and outcome variables. If the instrumental variable is not properly identified, the results may be biased or invalid.

Despite its challenges, the instrumental variable approach has been applied successfully in various fields, including economics, epidemiology, and political science. It provides a useful tool for researchers to estimate causal effects in situations where randomized controlled trials are not feasible or ethical.

In conclusion, both the design-based approach and the instrumental variable approach are important tools for investigating causal relationships in observational studies. Each approach has its strengths and weaknesses, and the choice of approach should be based on the specific research question and available data. By carefully selecting the appropriate approach, researchers can gain valuable insights into the causal relationships between variables of interest.