Difference In Difference Method: Description And Empirical Application
The Difference-in-Difference technique, abbreviated as “DiD”, is a research design originated in the field of econometrics which is used for the estimation of the causal effects. However, the basic logic of the technique has been used as early as the 1850s by John Snow and called the ‘controlled before-and-after study’ in some social sciences. Talking in the scenario of empirical economics, following non-experimental tool typically adopted for the estimation of the impact of governments' policy interventions on desired object/population. DiD is a quasi-experimental design in a canonical format which contains two groups and two time periods. Treatment group which has being enrolled in a program and the control group towards whom interventions haven’t been applied.
The methodology includes observing outcome for two groups for two time periods where one group(treatment) exposed to treatment in the second period but not in the first, while second group is not exposed to the treatment during either periods. In the case where the same units within a group are observed in each time period, the average gain in the second (control) group is subtracted from the average gain in the first (treatment) group in order to remove biases in comparison of two groups, which can be due to difference between individuals. By using longitudinal data from the design, changes in outcome over time compared and estimate the causal effect of interventions/treatments.
Difference in Difference estimation is a popular method to identify causal relationships and can be classified as reliable for its use of many years’ data. Moreover, it uses either level of data, individual and group, with ability of comparison starting at different levels of the outcome. DiD focuses on serially correlated outcomes and while estimating the effect of intervention, it takes into account time periods and compares outcomes before and after intervention has been incurred for both treatment and control groups.
However, in practise estimation is subject to severe serial correlation problems, which are ignored while researching and adopting Difference in Difference method; and there are three main factors which prove following criticism.
First of all, technique relies on the long-time series since it studies intervention effect by using longitudinal data, on average of 16. 5 years were found in Bertrand paper. Secondly, in typical scenario, dependent variables used in the technique are suffer from serial correlation. Lastly, intrinsic aspect of the DiD model, the treatment variable itself changes very little within a state over time.
Hence, following factors contribute to the fact that standard error could severely understate the standard deviation, and causal effect between two variables can be due to the other factors which has been completely omitted or hold incorrect functional form.
In the past 15 years, there has been a high focus amongst researches into how neighbourhood environment influences health. In 2004, Oakes in his literature documented a fact of selection-bias as a result of non-random nature of the neighbourhood selection process by individuals. Hence, one of the empirical application of DiD method is exploring problem such as self-selection bias in estimating the effect of neighbourhood environment on self-assessed health. Study by US Health and Retirement attempted to understand the direction of following bias by adopting longitudinal data set and focusing on relationship between health status and neighbourhood economic environment. Study concentrated on change in neighbourhood economic environment between 1992 and 2000 by comparing year 1992 which is before residential decision has been made and after the 1992–2000 when residential decisions were made. Change has been compared by applying baseline health status which clarifies if there is a bias and what is possible direction of it. Findings proved a strong association between neighbourhood environment and health by implying that those who live in disadvantaged neighbourhood reported worse health and vice versa.