The task is to write a research paper based on data from a randomized control trial where there are two different treatment interventions (T1, T2, to be defined by you). You can assume full compliance in the treatment. Your task will be to design the interventions and motivate these (i.e. the contribution to the literature).

You will be given a dataset (each student will receive a unique dataset) that contains N observations collected at two points in time from a target population (to be defined by you). The dataset will include:

•    Background characteristics such as education and gender collected at baseline.

•    Two dummy variables indicating whether the observation is in treatment 1, in treatment 2, or control (=0).

•    Four outcome variables collected at endline. These outcome variables are ‘standardized’ to have a normal distribution centred on 0, with mean 1. As with the treatments (T1 and T2) and the program (P), you decide the meaning of outcomes Y1, Y2, Y3, Y4.

Note: suppose your beta_1 = 0.2, where beta_1 is the coefficient estimate in a regression that includes T1 and where the outcome is  Y3. This means that “intervention T1 increases Y3 by 0.2 standard deviations”. I will discuss in class how to interpret this.

You imagine that you collected this dataset as part of your RCT, and analyze whether your evaluation has internal validity, and analyze the impact of your interventions on the outcomes of interest.

For example, you might decide that your treatment interventions are two different ways of encouraging people to participate in a job training program. Another example is that your program is university education, entrepreneurship training, an environmental program etc. You have free license to choose. You can be creative!

The only condition is that your program must betaking place in a low or middle income country.

Examples of dependent variables Y can be educational attainment, reduced conflict, increased earnings, improved recycling, etc.

Max. 3,500 words, including title page and abstract, excluding references and tables1.

1.5 spacing, 12 font, Font: Times New Roman

Each page should be numbered at the bottom, and follow the structure outlined below. You will also need to submit your Stata do-file.

Grading:

A.          Abstract – 5 points

B.           Introduction and literature review – 25 points

C.           Intervention and context – 15 points

1  Please put all tables in an appendix, and refer to the corresponding table in the text.

D.          Evaluation design and data collection – 10 points

E.           Data: Descriptive Statistics and Internal Validity – 10 points

F.           Results – 25 points

G.          Concluding remarks - 10 points

Impact Evaluation structure proposal

Please follow this structure:

A.   Abstract

Please summarize the research question, the experimental design and intervention, and key findings.

B.   Introduction

The introduction should include:

•    A brief summary of the societal challenge motivating the this study;

•    A brief summary of experimental design and intervention (details follow in the section below), and key findings (details follow in the results section);

•    An explanation of how this study contributes to the relevant literature (important!).

C.    Intervention and context

Please describe the project intervention, or interventions in case of multiple treatment arms. Make sure to describe the treatment arm(s), including a description of the intervention(s), the target beneficiaries, where the project was implemented, over how long, and provide some information about the local context that is relevant to the design, citing relevant data sources.

D.   Evaluation design and data collection

Please describe the evaluation design and data collection.

Make sure to describe the random assignment (e.g. are people, classrooms, businesses, communities, etc. being randomly assigned), sample size, timeline of data collection and intervention, and describe the data that has been collected.

While the data set contains background characteristics on the sample of participants (e.g. education, age, gender), the outcome variables Y1, Y2, Y3, Y4 are not yet defined. You will need to define what outcomes these represent, depending on your topic of choice. Please be specific in defining your outcome variables.

E.    Data: Descriptive Statistics and Internal Validity

First, present a balance table using the baseline data, and test for significant differences between the control and treatment groups in background characteristics.

Second, calculate the rate of outcome variable-specific attrition between baseline and endline, separately for each outcome variable Y.

Using OLS estimation findings, discuss whether there is evidence of selective attrition.