关键词 > Stata/Econometrics代写
Econometrics: Endogeneity and instrumental variables 2022/2023
发布时间：2023-01-03
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Econometrics: Endogeneity and instrumental variables
2022/2023
|
To overcome the second module of Econometrics (20% of final mark), the students, organized in pairs, must correctly carry out the following tasks. Deadline for submission: 2nd January 2023 (Aula Virtual / Submission. Endogeneity), attaching the following items: • The constructed cross-sectional dataset (in Excel, and Stata or R) • The PDF report with the answers. • The code (Stata do-file or R script). |
Select one of the following topics and, considering the suggested data source in each case, collect and build in Excel a cross-sectional dataset for a group of countries of your interest (minimum 25 countries) in a certain moment of time. Each team of students (max 2 person) should choose a different topic (please, register as soon as possible the chosen topic and the name of the team members in thisshared link).
TOPIC 1. Final consumption per capitai = f(GDP per capitai , other factor(s)i) for i = 1, 2, … , n countries.
• Y. Dependent variable:Households and NPISHs Final consumption expenditure per capita (constant 2015 US$)in 2015.
• X1. Main explanatory variable, possibly endogenous:GDP per capita (constant 2015 US$)in 2015.
• X2. Other explanatory factor(s) of your choice (not compulsory) in 2015.
• Zj. Instrumental variables:Gross capital formation (constant 2015 US$)in 2015, and the own lagged values of GDP per capita in 2014, 2013 and 2012.
• Data source:https://data.worldbank.org/
TOPIC 2. Energy use per capitai = f(GDP per capitai , other factor(s)i) for i = 1, 2, … , n countries.
• Y. Dependent variable:Energy use (kg of oil equivalent per capita)in 2014.
• X1. Main explanatory variable, possibly endogenous:GDP per capita (constant 2015 US$)in 2014.
• X2. Other explanatory factor (not compulsory):Industry (including construction), value added (% of GDP)in 2015.
• Zj. Instrumental variables: the own lagged values of GDP per capita in 2013, 2012 and 2011.
• Data source:https://data.worldbank.org
TOPIC 3. Life expectancyi = f(GDP per capitai , alcohol consumption per capitai) for i = 1, 2, … , n countries.
• Y. Dependent variable:Life expectancy at birth, total (years)in 2015.
• X1. Main explanatory variable, possibly endogenous:GDP per capita (constant 2015 US$)in 2015.
• X2. Other explanatory factor (not compulsory):Alcohol consumption per capitain 2015.
• Zj. Instrumental variables: the own lagged values of GDP per capita in 2014, 2013 and 2012.
• Data source:https://data.worldbank.org/
TOPIC 4. GDP per capita growthi = f(0fficial Development Aid received per capitai , other factor(s)i) for i = 1, 2, … , n developing countries.
• Y. Dependent variable:GDP per capita growth (annual %)in 2015.
• X1. Main explanatory variable, possibly endogenous:Net ODA received per capita (current US$)in 2015.
• X2. Other explanatory factor (not compulsory):Industry (including construction), value added (% of GDP)in 2015.
• Zj. Instrumental variables: the own lagged values of the Net ODA received per capita (current US$) in 2014, 2013 and 2012.
• Data source:https://data.worldbank.org/.
TOPIC 5. GDP per capita growthi = f(Natural resources rentsi, other factor(s)i) for i = 1, 2, … , n countries.
• Y. Dependent variable:GDP per capita growth (annual %)in 2015.
• X1. Main explanatory variable, possibly endogenous:Total Natural Resources Rents (% of GDP)in 2015.
• X2. Other explanatory factor (not compulsory):Labor force, total, in 2015.
• Zj. Instrumental variables: the own lagged values of the Total Natural Resources Rents (% of GDP) in 2014, 2013 and 2012.
• Data source:https://data.worldbank.org/.
TOPIC 6. GDP per capitai = f(Expense (% of GDP), other factor(s)i) for i = 1, 2, … , n countries.
• Y. Dependent variable:GDP per capita (constant 2015 US$)in 2015.
• X1. Main explanatory variable, possibly endogenous: Governmentexpense (% of GDP)in providing goods/services in 2015.
• X2. Other explanatory factor (not compulsory):Industry (including construction), value added (% of GDP)in 2015.
• Zj. Instrumental variables: the own lagged values of Expense (% of GDP) in 2014, 2013 and 2012.
• Data source:https://data.worldbank.org/.
TOPIC 7. Ifyou wish to propose your own topic (using data from the World Bank), please write to the email:[email protected]s.
Example of cross-sectional dataset in Excel (topic I):
|
i |
Country |
Y |
X1 |
Z1 |
Z2 |
Z3 |
Z4 |
|
1 |
Austria |
23283.608 |
44195.818 |
90930239902.460 |
44245.169 |
44299.378 |
44549.882 |
|
2 |
Belgium |
21053.226 |
41008.297 |
109240420185.156 |
40421.421 |
39970.317 |
39975.574 |
|
3 |
Denmark |
25089.143 |
53254.856 |
62445104478.215 |
52404.764 |
51831.798 |
51567.040 |
|
… |
… |
… |
… |
… |
… |
… |
… |
|
50 |
Spain |
15055.102 |
25742.369 |
227119507736.719 |
24772.341 |
24361.257 |
24635.176 |
where
• Y: Dependent variable: Households and NPISHs Final consumption expenditure per capita (constant 2015 US$) in 2015 .
• X1. Explanatory variable (possibly endogenous): GDP per capita (constant 2015 US$) in 2015 .
• Z1. Instrumental variables: Gross capital formation (constant 2015 US$) in 2015 .
• Z2. GDP per capita (constant 2015 US$) in 2014.
• Z3. GDP per capita (constant 2015 US$) in 2013.
• Z4. GDP per capita (constant 2015 US$) in 2012.
Once constructed the dataset, answer the following questions, employing the software STATA (or R) for the statistical analysis and elaborate a PDF report with your answers and results. Please, if applicable, do not forget indicating the null and alternative hypotheses when a test is conducted and interpreted.
1.Define the variables of the constructed dataset (indicating their measurement units and statistical source).
2. Evaluate the relationship between the dependent variable Y and the explanatory variable of interest X1, using the scatter plot and the coefficient of correlation.
3. Choose the most adequate econometric specification (level-level, log-log, log-level, level-log) for evaluating the possible linkage between the dependent variable and the explanatory variable(s). Please, remember that the scatter plot could be useful for this purpose. Use the adequate econometric specification from now on.
4. Estimate by Ordinary Least Squares (OLS) the chosen econometric specification. Later, present and interpret the resulting Sample Regression Function and the main outcomes (that is, interpret the estimated coefficients and statistical significance).
5. If applicable, identify and briefly explain the potential source of endogeneity in your specification.
6. Estimate by Two-stages Least Squares (2SLS) the econometric specification, using instrumental variables. Please, try to include more instruments than potential endogeneous variables to be able to evaluate their adequacy. Later, present and interpret the resulting Sample Regression Function and the main outcomes (that is, interpret the estimated coefficients and statistical significance).
TIP: The own lagged values (two or three time periods) of the endogenous variable can be employed as instruments. See,
for example: Reed, W. R. (2015). On the practice of lagging variables to avoid simultaneity. Oxford Bulletin of Economics and Statistics, 77(6), 897-905.
7. Once estimated the model by 2SLS, make sure that employed instruments are suitable:
7.1. ¿Are the instrumental variables relevant? Regress by OLS the possibly endogenous variable X1 on exogenous variables
and instrumental variable(s). Then, test the (joint) significance of instrumental variable(s).
7.2. Are the instrumental variables uncorrelated with the disturbance process? Perform and interpret the Sargan test.
TIP: If lagged values of the endogenous variable are employed as instruments, try to select the adequate number of lags to guarantee relevance and exogeneity of instruments.
8. Run and interpret the endogeneity tests purposed by (A) Hausman and (B) Durbin-Wu-Hausman.
9. Are the errors homoskedastic? If this is not the case, use the Generalized Method of Moments (GMM) estimator with heteroskedasticiy-consistent standard errors. Later, present and interpret the resulting Sample Regression Function and the main outcomes (i.e., estimated coefficients and statistical significance).
10. According to the obtained test outcomes, which estimator (OLS, 2SLS, or GMM) would be the most adequate? Why?
