ECON0019: QUANTITATIVE ECONOMICS AND ECONOMETRICS 2022
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
ECON0019: QUANTITATIVE ECONOMICS AND ECONOMETRICS
EMPIRICAL PROJECT: LATE SUMMER ASSESSMENT 2022
Instructions
The mark for the empirical project is worth 20% of your total mark for the module.
Please follow these instructions so that we can ensure anonymity in marking and ensure compliance with UCL assessment policies. We will only be able to give you credit for your project if you follow these instructions. If the instructions are not followed, you will receive a mark of zero.
1. All answers must be uploaded via Turnitin by 12pm on August 31, 2022.
2. All marking on Turnitin is anonymised. Do not put your name or student number or group name anywhere on your submitted answer — either in the document or in the file name.
3. Submit the project with the submission cover sheet on the first page. Put your candidate number on the submission coversheet. Your candidate number is NOT you student number! Use the candidate number from this year—it is not the same as last year.
4. You should submit one PDF or Word document that includes: you answers and expla- nations in the main text (including tables and figures, if any), as well as an appendix with your code producing these results. If you use software other than Stata, you should state which programme was used. You may optionally include raw statistical output (e.g. Stata log-file) after the code but such output does not substitute for your answers and explanations.
5. Your answers should be no more than 800 words, including footnotes but excluding tables, figures, and the code appendix. You must state the number of words at the top of the first page of your submission.
6. Submissions will be checked for plagiarism. By submitting this assessment, you pledge your hon- our that you have not violated UCL’s Assessment Regulations which are detailed in https:// www.ucl.ac.uk/academic-manual/chapters/chapter-6-student-casework-framework/section- 9-student-academic-misconduct-procedure, which include (but are not limited to) plagia-
rism, self-plagiarism, unauthorised collaboration between students, sharing my assessment with another student or third party, access another student’s assessment, falsification, contract cheat- ing, and falsification of extenuating circumstances.
7. Please make sure to allow sufficient time should problems arise with Turnitin. Check the submis- sion inbox for confirmation that your essay has been submitted. Once your submission has been accepted you will return to the ‘My Submissions’ tab where you will be able to see the details of your submission. If your submission is not confirmed for some reason, or you are having issues uploading the document, get in touch with ISD ([email protected]) as soon as possible to figure out what the problem might be.
You will be awarded a mark of 0% or Grade F if you (1) do not attempt the summative assessment component or (2) attempt so little of the summative assessment component that it cannot be assessed. Please check the UCL Academic Manual (Section 3. 11) for information on the consequences of not submitting or engaging with any of your assessment components .
QUESTION:
In “Gasoline Taxes and Consumer Behavior” (American Economic Journal: Economic Policy, 2014), Shanjun Li, Joshua Linn, and Erich Muehlegger study the effects of gasoline taxes and the price of gasoline excluding such taxes on gasoline (i.e., petrol) consumption. They are particularly interested in whether taxe changes have the same or bigger effects on consumption compared with price changes unrelated to taxes.
To study this question, they collect annual panel data on taxes, prices, and consumption from 1966–2008 (with t denoting years) for 48 states of the United States (with s denoting the states). They decompose the dollar retail price Pst into the price excluding tax Est and the per-gallon tax Taxst:
Pst = Est + Taxst
or, in logs,
log Pst = log Est + log Tst , where Tst = /1 + 、 .
The main specification they are interested in is
log Qst = β 1 log Tst + β2 log Est + γs + δt + ust , (*)
where Qst is per capita gasoline consumption by state (s) and year (t), β 1 and β2 capture how the demand for gasoline depends on taxes and on the tax-exclusive price, respectively. Next, γs are the state “fixed effects” — time-invariant demand shifters, implemented by including state dummies in the equation. Similarly δt are year fixed effects, capturing national shocks to gasoline demand and implemented by including year dummies.
The Stata data file ECON00192022LSA .dta contains observations on the variables of interest. Specif- ically:
◆ year — year of observations (1966–2008);
◆ Iyear* — year dummies;
◆ state — name of the state (and fips is the numeric code of the state);
◆ Istate* — state dummies;
◆ lngca: (log) annual gasoline consumption per adult;
◆ lngpinc: (log) retail (i.e., tax-inclusive) gasoline price (denoted log Pst);
◆ l*lngpinc: lags for lngpinc;
◆ lngp: (log) tax-exclusive gasoilne price (log Est below);
◆ lntr: the tax-related component of the (log) retail gasoline price (log Tst);
◆ lntax oil: the instrument Zst introduced below;
◆ l1lntax oil: the first lag of Zst;
◆ tax up: dummy variable for state taxes increasing in a given year;
◆ fd hsgrad: annual change in fraction of adults graduating in high school.
We simplify some of the analysis from the paper. In all questions (except question 2, as explained there) use standard errors which are robust not only to heteroskedasticity but also to serial correlation of the error term — “clustered standard errors.” To implement these in Stata, use option cluster(state) instead of robust, where state indicates that all observations for a given state across all years belong
to the same “cluster” in which error terms may be correlated.
Some Stata hints:
◆ Command test allows you to compute F-statistics and perform two-sided tests on (single or multiple) coefficients or their linear combinations. Type help test to get more details for how this command works.
◆ To plot estimated coefficients after running a regression, you can use the command coefplot, vertical xlabel(,labsize(small) ang(90)) after installing the programme coefplot using ssc install coefplot. You can add other graph options to make the graph look better (e.g. label the y axis).
Answer the following questions:
1. Estimate the following regression:1
log Qst = 6t + ust .
How do you interpret the coefficients 6t? Plot the estimates for 6t . In 1973, an oil embargo by the Organisation of Arab Petroleum Exporting Countries led to the First Oil Crisis, and in 1979,
the Iranian Revolution drove down oil production leading to the Second Oil Crisis. How is this reflected in the graph? Explain your answer.
|
2. Estimate model (*) by OLS with clustered standard errors and without clustering (i.e., with heteroskedasticity-robust standard errors). Report the OLS coefficient estimates βˆ1 and βˆ2 with three significant digits. Report standard errors and compare them to each other. How do you interpret the difference? Which standard errors do you prefer? (Use clustered standard errors for the rest of the empirical project.)
3. You conjecture that the log(Tst) and log(Est) affect log(Pst) in a non-linear fashion and so decide to include quadratic terms and an interaction term in the model,
log(Qst) = β1 log(Tst)+β2 log(Est)+β3 (log(Tst))2 +β4 (log(Est))2 +β5 [log(Est).log(Tst)]+γs+δt+ust .
(#) Estimate the model (#) by OLS. Report the (OLS) coefficient estimates βˆ1 , ..., βˆ5 with three significant digits, along with their standard errors. Write the equations for the (estimated) par- tial effects of log(Tst) and log(Est), respectively. Compute the average partial effect (APE) of each of these variables. Are these partial effects different from those from model (*)? (You do not need to test whether the difference, if any, is statistically significant.)
4. Test the linear model (*) against the quadratic alternative (#). Conclude.
5. In arguing that Taxst is exogenous, the authors examine whether changes in state taxes are related with changes in demographic and political variables in the state. The variable tax up records whether there was an increase in state gas taxes in a given year relative to the previous year. Estimate a logit model of tax up on fd hsgrad, the change in the proportion of high- school graduates in the state. (Do not include any fixed effects.) Test the null hypothesis that the coefficient on fd hsgrad in your logit model is zero. Is there evidence that changes in the proportion of high-school graduates in the state are related to increases in taxes?
6. What is the partial effect at the average (PEA) of the change in the proportion of high-school graduates on the likelihood of a state gas increase? Report this PEA with three significant digits along with its standard error and provide a verbal interpretation.
The authors are further worried that log Est, which is determined by the equilibrium of time-varying state-level demand and supply, may be endogenous in equation (*). This would also make log Tst potentially endogenous, since Est enters its calculation. They set out to instrument both endogenous explanatory variables. For log Tst they come up with an instrument:
Zst = log /1 + 、 ,
where Crudet is the price of imported oil in year t (the same across all states). Since Crudet is not determined in the regional gasoline market, it is not subject to the simultaneity problem, helping exogeneity of log Tst .
7. A colleague suggests that having more instruments is always better, and you can use as second instrument the first lag of Zst, denoted Zst − 1 : if Zst is exogenous, it is very likely that Zst − 1 is exogenous as well. Do you agree with the colleague’s view and their proposal? Estimate model (*) by two-stage least squares, using Zst and Zst − 1 as instruments for log Tst; view log Est as exogenous for now. Report the corresponding 2SLS coefficient estimates β˜1 and β˜2 with three significant digits, along with their standard errors.
8. Check if there is a weak instruments problem in this 2SLS estimation. Report which statistic you used, what value it takes, and the appropriate critical value. Check if this statistic is higher or lower if only Zst was used as a single instrument. Discuss why the difference between the values of this statistic arises.
9. Now that you have two exogenous instruments, a friend recommends to relax the exogeneity assumption for log Est and use log Zst and log Zst − 1 as instruments for log Tst and log Est . Re- port the corresponding 2SLS coefficient estimates βˇ1 and βˇ2 with three significant digits, along with their standard errors. Compare these estimates and their standard errors with those you obtained previously by OLS and 2SLS, and discuss the difference. (Note: You do not need to conduct formal statistical tests for the question.)
10. To examine the dynamics of gasoline prices, estimate the following auto-regressive (AR) model: log Pst = α0 + α1 log Ps,t − 1 + α2 log Ps,t −2 + α3 log Ps,t −3 + γs + δt + ust .
Test the (joint) null hypothesis that α 1 = α2 = α3 = 0. What does this imply about the persis- tence of shocks to gasoline prices?
2022-08-23