PPGA 503 – Take Home Final 2022
Due: 11:59 pm on Dec 15 This is the take home component of your PPGA 503 Final Exam. You may brainstorm with your classmates when beginning your work, but all of the analysis and write-up are to be completed individually. Please upload three deliverables to Canvas:
• Annotated .do file you used to complete your analysis
• Technical write-up
• 1-page memo For the take home portion of your final, you will again use a modified version of the World Bank’s
Indonesia Database for Policy and Economic Research (INDO-DAPOER) dataset. You shouldn’t need it, but you can access the background material on the dataset in the following places: here and here. The dataset and a list of variables are available on Canvas (under final exam). As you know, the dataset includes a significant number of variables covering health, education, governance, economic, development, and natural resource attributes at the district level in Indonesia. There are slightly over 500 districts in one of two types: kota (city) are more urbanized, while kabupaten (regency) are typically more rural. Provinces, of which there are just under 40, form the meso (middle) tier of government. You previously focused on health outcomes. For this assignment, you will choose one of the following three prompts, each with a different DV. You will complete the necessary data management and analysis to answer your chosen prompt. We are largely leaving the analysis to you, though the write-up instructions (together with HW 4 and 5) will be useful as a guide.
Prompts (please choose one of the following three) 1. Educational performance: Educational performance is critical for development, as it is directly empowering for local populations, and has all manner of spillover effects (for instance in attracting new investment and economic opportunity). Our dataset has a district-level variable called Average
National Exam Score: Senior Secondary Level, which captures performance in a nationally-administered exam. The score is out of 100 points. You are asked to provide insights (1) on variation in the exam performance (ie, why some districts do well and others do not), and (2) on how (based on those insights) districts might improve their performance. Remember that some district-level attributes are fixed (a district can’t change its location or demographic structure, for example, but it can build more schools, raise/lower taxes, etc…). 2. Inequality: Inequality receives a lot of attention in popular circles and among policy makers, as it has important practical and normative implications. The Gini Index is a frequently-used measure of inequality (higher values denote higher levels of inequality. Our dataset includes district-level Gini coefficient. You are asked to provide insights (1) on variation in the Gini coefficient (ie, why some districts are highly unequal, while others are more equal), and (2) on how (based on those insights) districts might reduce inequality levels. Remember that some district-level attributes are fixed (a district can’t change its location, demographic structure, or natural resource endowment, for example, but it can make a number of potentially relevant policy changes). 3. Household access to electricity: Access to electricity is seen as important for quality of life and for empowering populations, as it has benefits for education, health, and productivity. Electrification (Household Access to Electricity: Total (in % of total households)) varies considerably across Indonesian districts. You are asked to provide insights (1) on variation in household electrification rates (ie, why some districts have high rates and others low rates), and (2) on how (based on those
insights) districts might increase electrification rates. Remember that some district-level attributes are fixed (a district can’t change its location, demographic structure, or natural resource endowment, for example, but it can make a number of potentially relevant policy changes). Your technical write-up should answer the following questions. You are writing this for the instructional team, so you are free to use precise and technical language. 1. Theoretical/Conceptual a. Once you have selected your prompt, think carefully about the outcome you are trying to explain. From a theoretical/conceptual perspective (don’t think about specific variables yet), what do you think is important for understanding variation in the DV? In the real world, of course, everything matters. But some things matter more than others. Here we’d like you to focus on the most important factors (or groups of factors). Please write one paragraph summarizing, from a theoretical/conceptual perspective, what you believe is important for explaining variation in your DV. Don’t go overboard (we don’t want 30 distinct factors), but do try to be reasonably comprehensive. You might also distinguish conceptually between core and peripheral (or subsidiary) factors. b. Now let’s think about conceptualizing and operationalizing those factors. Look at the variables available in the dataset. Are you able to capture all of the factors you identified in 1a with the available variables? How comfortable are you with those variables? In other words, do you think the available data can effectively capture the factors you’ve identified, or do you have concerns? Recall that you can always construct new variables based on existing ones. Please write one paragraph that discusses how well you think the available data can capture the theoretically relevant factors you previously identified, making sure to note specific concerns. 2. Models a. Please do any necessary data preparation (re-scaling, renaming, constructing new variables, etc) and then start constructing your model. We’d like you to proceed in three stages. The first (model 1) should have the core independent variables (somewhere between 2 and 4). The second (model 2) and third (model 3) should make incremental refinements. This will primarily be the inclusion of additional variables, but you may also include other refinements such as interaction effects, functional form transformations, etc, if you think they are
necessary (in other words, don’t include them just for the sake of being fancier!). Please write
at least four paragraphs: in the 1st, briefly describe any data preparation you did. In the 2nd, discuss how useful the models are overall, being sure to address how they change through the refinements; in the next paragraph(s), interpret with some precision the independent variables, focusing on both statistical and substantive importance (you may use more than one paragraph if you feel necessary); in the final paragraph, please briefly discuss any important decisions you made in the analysis, and note any remaining concerns you have. b. Please look for and address remaining issues with the models (some of which you may have noted in the previous paragraph). This might include things violations of the Gauss Markov assumptions, the strong effect of outliers, etc. If you make any corrections, estimate the regression again and include the results as model 4. Please write one paragraph that discusses what you did and whether it changed your overall conclusions in any meaningful way. c. Please construct a (nicely edited) regression table containing all four models, and include it in your write-up. Remember outreg2 (or similar) to save you from doing this manually! 3. Reflections
a. Now that you have completed the full analysis, step back to reflect on the findings. Do they support or contradict your initial expectations? Is there anything you find surprising? Are there any major omitted variables that you haven’t accounted for yet? Please discuss these (or related) issues in one or two paragraphs. b. Finally, we care about causality because it is key to changing outcomes via policy. Think critically about your findings, especially between key IVs and the DV. Are we picking up correlations? Causations? An endogenous relationship? Think about these issues, and then discuss in one paragraph. Finally, complete your 1-page memo. Think of this as a briefing for a policy maker, addressing the questions posed in the prompt. You should provide basic contextual information (data source, descriptive statistics, etc), an overview of your conclusions, and a short discussion of implications, particularly in terms of potential policy interventions. You should write this for the policy maker (not for the instructional team), so be sure that your language is precise but accessible, and that the analysis is grounded in evidence. You may structure the memo however you’d like (including using visualizations), but you are limited to one page (to be clear, one side, not both!).