ECON0019: QUANTITATIVE ECONOMICS AND ECONOMETRICS

LATE SUMMER ASSESSMENT 2024

Instructions

The mark for the empirical project is worth 20% of your total mark for the module.

Pleasefollowtheseinstructionssothatwecanensureanonymityinmarkingandensurecompliance

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 Moodle by 9am on August 29, 2024.

2. All marking is anonymised. Do not put your name or student number anywhere on your

submitted answer — either in the document or in the file name.

3. You should submit one PDF or Word document that includes: your answers and expla-

nations in the main text (including tables and figures, if any), as well as an appendix with

your code producing the 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.

4. 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.

5. Submissionswillbecheckedforplagiarism. Bysubmittingthisassessment, youpledgeyourhon-

ourthatyouhavenotviolatedUCLAssessmentRegulationswhicharedetailedinhttps://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) plagiarism, self-plagiarism, unauthorised collaboration

between students, sharing my assessment with another student or third party, access another

student’s assessment, falsification, contract cheating, and falsification of extenuating circum-

stances.

6. PleasemakesuretoallowsufficienttimeshouldproblemsarisewithTurnitin. Checkthesubmis-

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.

7. According to UCL’s AI guidelines, this assessment falls under Category 1: AI tools cannot

be used. The use of AI tools that exceeds that permitted in the assessment brief constitutes

Academic Misconduct.

ECON0019 LSA 1 TURN OVER

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 “Competition and Innovation: An Inverted-U Relationship” (Quarterly Journal of Economics,

Vol.120, No.2, 2005), Philippe Aghion, Nick Bloom, Richard Blundell, Rachel Griffith and Peter

Howitt study the relationship between competition and innovation. This question is based on the

specifications in the article, which use measures for both variables and investigates their empirical

association. The innovation measure used in the study is (essentially) the number of patents issued in

an industry during a given year over the period 1973 to 1994. The unit of observation is the industry

at each (available) year and there are 17 industries in the data. To register the degree of competition

in an industry, the authors use one (1) minus a measure called the Lerner index. An industry in

perfect competition would have the index equal one. Lower values for the index register deviations

from perfect competition.

The main variables in the dataset empirical proj 2024.dta are:

Variable Code Variable

sic2 Industry Code

patcw Citation weighted patents

Lc Competition

yr∗ Year Dummies

1. Provide summary statistics for the variables above and obtain OLS estimates for the following

regression:

patcw = α +α Lc +α Lc2 +year dummies+u , (1)

jt 0 1 jt 2 jt jt

where patcw and Lc are the citation weighted patents and competition measure, respectively,

jt jt

in the jth industry in year t, j = 1,...,17 and t = 1,...,22. Throughout this project, you should

use heteroskedasticity robust standard errors. Are the coefficients on Lc and Lc2 jointly

jt jt

significant? How do you interpret them? What is the Average Partial Effect (APE) of Lc

jt

on patcw ? (Note that this is equal to the Partial Effect at the Average (PEA).) Based on

jt

the estimates, which level of competition produces the highest level of innovation? What is the

marginal effect of Lc for an industry where Lc = 0.5? What is the marginal effect of Lc for

jt jt jt

an industry where Lc = 1 (i.e., perfect competition)?

jt

ECON0019 LSA 2 CONTINUED

2. Suppose you want to estimate the probability that an industry issues a positive number of

patents in a given year. To do that, create a dummy variable Y that records whether industry

jt

j issued at least a positive number of patents in year t or not and run the following model:

Y = 1(β +β Lc +β Lc2 +year dummies+u ≥ 0)

jt 0 1 jt 2 jt jt

where u ∼ N(0,1). What are the PEA and APE for Lc ? Why are they different while the

jt jt

PEA and APE for equation (1) are equal?

3. Note that the patcw is zero for 12.99% of the observations. Estimate a Tobit model where

jt

patcw∗ = γ +γ Lc +γ Lc2 +year dummies+u (2)

jt 0 1 jt 2 jt jt

and patcw = patcw∗ if patcw∗ > 0 and patcw = 0, otherwise. Are the coefficients on Lc

jt jt jt jt jt

and Lc2 jointly significant? What is the (unconditional) Average Partial Effect (APE) of Lc

jt jt

on patcw ? How does this compare with the APE obtained in item (1)? What is the marginal

jt

effect of Lc for an industry where Lc = 0.5? What is the marginal effect of Lc for an

jt jt jt

industry where Lc = 1 (i.e., perfect competition)? What is the Average Partial Effect (APE)

jt

of Lc on the probability that patcw > 0? How does this compare with the APE obtained in

jt jt

item (2)?

4. Notethatthedatasetisapanelofindustries. Writeupapaneldataversionof(1)with(industry)

fixed effects (and maintaining year dummies). Provide an interpretation of the fixed effects in

this particular context. Estimate the fixed effects model; how do the coefficient estimates on

Lc and Lc2 compare with those from the OLS regression in Question 1? What is the Average

jt jt

Partial Effect (APE) of Lc on patcw ? What is the marginal effect of Lc for an industry

jt jt jt

where Lc = 0.5? What is the marginal effect of Lc for an industry where Lc = 1 (i.e.,

jt jt jt

perfect competition)?

5. The authors point out that high levels of innovation may reduce competition and this would

make Lc endogenous. Does the panel data model you estimated in the previous question allow

jt

you to control for this type of endogeneity? Discuss.

To address this problem, the authors employ instrumental variables that reflect policy changes

affecting competition. Those policies are the Thatcher era privatisations, the EU Single Market

ProgrammeandtheMonopolyandMergerCommissioninvestigationsthatresultedinstructural

changes in certain industries. Those variables in the data set are listed below:

ECON0019 LSA 3 TURN OVER

Variable Code Variable

SMPhighD Single Market Programme high impact

SMPmedD Single Market Programme medium impact

car Car Industry

per Periodicals Industry

brew Brewing Industry

tele Telecoms Industry

phar Pharmaceuticals Industry

text Textiles Industry

raz Razor Industry

steel Steel Industry

ord Ordnance Industry

rd yUSA Industry R&D/Y USA

rd yFRA Industry R&D/Y France

tfpUSA Industry TFP USA

tfpFRA Industry TFP France

imp yUSA1 Industry Imports/Y USA

imp yFRA1 Industry Imports/Y France

exp yUSA1 Industry Exports/Y USA

exp yFRA1 Industry Exports/Y France

muUt Markup USA

liUSA1 Output minus variable costs over output USA

muFt Markup France

liFRA1 Output minus variable costs over output USA

Provide the TSLS estimates for the coefficients in the following version of regression (1):1

patcw = α +α Lc +year dummies+u . (3)

jt 0 1 jt jt

Are the instruments sufficiently strong? Test for endogeneity of Lc using regression-based test

jt

covered in class.

6. The data for each industry can also be viewed as time series data. Plot the time series of

citation-weighted patents for the pharmaceuticals industry. Plot the autocorrelation function

for the same time series. Is the data persistent? Does it follow a trend? Estimate an AR(1)

1Item(4)demonstratesthat(industry)fixedeffectsareimportant. Ideallyyouwouldliketoestimatethepaneldata

model there accounting for endogeneity in Lc and Lc2 . Two problems arise: ( i ) applying TSLS in the presence of

jt jt

(industry)fixedeffectsand(ii)thefactthattheendogenousvariableshowsupinLc andinLc2 . Thisispossible,but

jt jt

is a topic for more advanced courses.

ECON0019 LSA 4 CONTINUED

model for the series. Are you worried about stationarity? If the other time series exhibit similar

behaviour, are you concerned about the standard errors used in your other answers? Why?

ECON0019 LSA 5 END OF PAPER