CHAPTER 1

INTRODUCTION

Background of the Study

Since the beginning of civilization, there have been a lot of strategies

used in developing certain areas. There have been a lot of policies/ program

initiated to improve and satisfy human needs. The success of the intervention

is also the success of the government or those organizations that has a goal

in improving certain areas. The question of interest when investigating the

effect of some policy intervention is to evaluate the consequences of the

policy on the units or populations subject to this policy. We want to find out if

those populations are better off after they have been subject to the policy

(Miquel, 2003). But how do determine if the interventions are really effective?

Evaluation helps to answer this question. The main purpose of evaluation is to

improve the quality of a program or a project by identifying its strengths and

weaknesses. Evaluation is an important part of intervention.  Extension

programs, no matter how large or small, need to be reviewed or assessed to

see if they accomplished the stated objectives.  Through evaluation

processes, this would able to find how the intervention benefitted the

community (Worthen et al, 1987).

There have been a lot of researchers whose objective is to evaluate or

measure the effects of a project. Most researchers in every field is facing

problems regarding the estimation of the effect of a certain project. One of the

difficulties is the methods being used. They could not estimate the effect in
just one observation, it always has a complication or that they used multiple

methods in estimating one problem.

The project evaluation is apparent in different fields such as medicine,

physics, biology, psychology and social sciences like sociology. For example,

we might wish to estimate the effect of a welfare reform on labor force when

training programs are organized by the government for the labor market as

well as to those unemployed (Matlack, 1993). In medicine for instance, we

wish to estimate the change in health care policy and how it affects the health

outcomes of younger and older people (Wooldridge, 2011).

We are in the era of accountability and the demand for project

evaluation is growing. A new method is proposed that could be used in

evaluation of a given intervention. It entails the collection of panel data or

cross sections, the Difference-in-Differences (DiD) estimator (Viscusi et al,

1995). DID integrates the advances of the fixed effects estimators with the

causal inference analysis when unobserved events or characteristics

confound the interpretations (Angrist and Pischke, 2008). Basically, DID

estimation involve 3 types of condition to be able to use to quantify the effects

of the intervention;

1.) Standard DID estimation. This is similar to fixed effect estimator but

DID measures less standard error compared to fixed effect estimator.

However, this could be used to pool cross section as long as the group

or area completes the requirement of the before and after the

intervention.
 2.) DID Estimation can also involve more than one group or observations

taken over several intervals of time, but this have features of random

selection.

3.) DID estimation can also extend to measure more than two

dimensions to create Difference-in-Difference-in-Differences (DDD)

estimation. This method allows one to measure the effect on different

groups (Lechner,2010).

Rationale of the Study

Despite the existence of other plausible methods based on the

availability of observational data for quasi-experimental causal inference (i.e.

matching methods, instrumental variable, regression discontinuity), DID

estimations offer an alternative to reach the unconfoundness by controlling

the unobserved characteristics and combining it with observed or

complementary information. Additionally, the DID is a flexible form of causal

inference because it can be combined with some other procedures, such as

the Kernel Propensity Score (Heckman et al., 1997) and the quintile

regression (Meyer et al.,1995).

Difference - in- Difference (DID) Estimation is useful for policymakers to

quantitatively see the true effects of the policy or program, it could be an

attractive choice as a research design in estimating the effect of an

intervention. The DID measures the effect of a treatment (independent variable)

on an outcome (dependent variable) by comparing the average change

overtime in the outcome variable for the treatment group to the average

changes overtime for the control group (Lechner, 2010). This methodology
needs data on, before and after the program where one has the opportunity to

select individuals randomly on treatment group and the control group (Matlack,

1993).

One advantage of DID usage in measuring the effects of intervention is

that, it does not require a broad knowledge about econometric principles with

regard to understanding its concept. It is easy to obtain the final estimates and

their standard errors when identifying the problem using its regression

formulation. Furthermore, the model could be easily extended to cover more

periods and more treatments, including continuous treatments and add

additional covariates without much further computational effort (Lechner, 2010).

The estimation procedure differs from other methods for three (3)

reasons. First, it does not require repeated observations for the same

individuals. It allows selection on unobservable characteristics and is less data

consuming (it does not require to observe all variables simultaneously

influencing the participation decision and the potential outcomes). The

proposed estimators are feasible under the data requirements for traditional

DID estimators when applied to repeated cross-sections. Second, it allows the

estimation of parsimonious parametric approximations to the average effect of

the treatment on the treated conditional on selected covariates of interest.

Finally, the framework can accommodate a multilevel treatment variable, which

is different treatment intensities (Abadie, 2005).

The DID is a powerful tool but is not extensively used because of the key

assumption, the “Parallel Path”, which posits that the average change in the

comparison group represents the counter factual change in treatment group if

there were no treatment. The parallel path assumption fails when the
treatment group and the control group cannot obtain the normal difference.

This means that whatever happened to the control group over time is what

would have happened in the treatment group in the absence of the

intervention (Gong, 2009). The failure of the parallel path assumption in every

investigation has a bias result. To solve this problem, the estimation requires

data from only two points in time and the results are robust to any possible

confounder. Another possible solution with this problem is to get more data

on the other time periods before and after treatment to see if there are any

other pre-existing differences in trends (Albouy,2004)

The problem is that changes in the outcome of interest might be

systematically different across states due to, say, income and wealth

differences, rather than the policy change. A more complex analysis than

either of the DID analyses which can be obtained by using both a different

state and a control group within the treatment state (Wooldridge, 2007).

Objective of the Study

The general objective of the study is to apply the Difference in

Difference (DID) estimation to determine the impacts of a project. Specifically

the study aims;

a) To use the DID estimation as a possible design in evaluating the

projects; and

b) To compare the results of the estimates obtained with other standard

procedures such as T-test or F-test.

DID estimation is a research design developed statistically to interpret

the results and to establish another strategy on evaluating the effect of an

intervention. This study gives somewhat different discussion compared to the

other textbooks in discussing the same estimation. Thus, this research is not

a substitute reference but this is a complement to the other researches

regarding research design.

This study was made to introduce DID estimation to economic students

as well as to the policy makers and policy analysts as an alternative method

to evaluate impact of project on a certain interest. This can also be used to

develop and explore the strength of the linear regression method. Take in

mind that DID estimation is another chain of linear regression analysis but has

different interpretation and different goals of interests. This is intended to

apply and use the DID estimation in economic analysis.

This will benefit students in different fields, policy makers and policy

analysts to have a guide in the new propose method to use for evaluating the

effect of intervention.

Scope and Limitation

This study will focus in overviewing Difference in Differences

Estimation strategy and gives some issues mainly using treatment effect

perspective and give some observational projects to be able to estimate the

effect. The sample data of the project will help to prove the effectiveness of
the propose method if the data will suit to the assumptions of DID. Thus, this

study limits its variable use to observe clearly the efficiency of the method.

Structure of the Study

Chapter 2 presents the related literature of the study. This tackles about the

issues and variable used to estimate the effect of the intervention.

Chapter 3 presents the Methodolgy of the study. These includes the

Digression and the assumptions of DID. Digression provides an overview in

the mathematical formulation of DID Estimation while the assumptions of DID

will focus to the usual violations of Gauss Markov Assumptions or the

Assumptions to estimate efficient multiple regression.

Chapter 4 presents the results and discussion of the study. These

includes the result the evaluation of the two project and the efficiency of DID

estimator for project evaluation.

Chapter 5 presents the Summary, Conclusion, recommendation and

the areas for further research of the study.