Faculte Des Hautes Etudes Commerciales De L'universite De Lausanne

FACULTE DES HAUTES ETUDES COMMERCIALES DE L'UNIVERSITE DE LAUSANNE

Final exam

Professeur :
Olivier Cadot / Matière :
Globalization and development / Session :
Fall 2014

• Duration: 2hours
• No documentation
• No calculator
• Exam is7 pages long
• Please return the exam questions withyouranswers
• Multiple-choice questions have only one correct answer, and thereis no penalty for wronganswers.
• If you are stuck on one question, do not panick; just move on to the next. Make sure youkeepsome time to re-readyouranswers at the end of the exam.

Name, first name / Matricule # / Seat #
Grade

Question 1

1. (10 points) Whatistrade in value added?In youranswer, pleasedefine value added, explainin whatsensetrade datais not comparable to GDP data in terms of value added, explainpreciselywhy and how trade data mayinvolve double counting (possiblywith an example if ithelps). Explainwhat type of data isneeded to calculated exports in value added. How didKoopman et al. (2011) solve the problem?

Trade in value addedistrade in which the value of exports iscalculated as gross value minus all intermediateconsumptions. Calculating exports thatwayreconcilesitwith GDP data.
Right now, trade data is not comparable with GDP data because GDP data isbased on value added (GDP isdefined as the sum of value added by all firms in the country) whiletrade data counts the gross value of each export transaction.
Double countingcomes in the followingway. Suppose that China exports an ipod. Most of the ipod’s components are imported by China (fromJapan, Korea or the US). Thus, they have already been counted in international trade data. If the ipodwassold in China, national accountswouldonly count the value added, not the value of the imported components. By contrast, if the ipodisexported, everythingiscounted.
In order to calculate the value-added content of exports, one would have to use eitherfirm data or input-output tables. However input-output tables typically do not distinguish at the sectoral levelwhere inputs come from (domestic or imported), sosome approximations have to beused. Koopman et al. (2011) combined input-output data withtrade data by product to come up with approximations.

1. (10 points) In whatsectors do youexpect to see the smallest/largestdifferencebetweengross exportsand exports in terms of value added? Explain and give an example of a country (or type of country) for which the differencedoesnotmattertoomuch, based on the sectoral composition of its exports. Hint: This isnot the place to talk about the ipod!

The differencebetweengross exports and exports in terms of value addedissmallest for productsusing few intermediateproducts, for instance mining or agricultural products. Therefore the differencedoes not mattertoomuch for countries, say in sub-SaharanAfrica or the Middle East, which export essentiallyprimaryresources (oil, minerals, or agricultural products).
By contrast, the differenceislargest in the case of manufacturedproducts; typically, the more complexis the production process—say, for electronics, machinery, transportation equipment, and the like) the more it uses imported components. It isalsotrue of clothingproducts.

1. (10 points) For what type of countries is the differencebetweengross exports and exports likely to mattermost? Giveexamples and explain.Wouldstiffrules of origin in a free-trade agreement have anyeffect on this?

It is for countries exporting products after final assembly, importing most or all of the components. Prominent examples include China or Mexico. The reason is that the local value added is minimal.
Yes rules of origin forcing the country exporting a final (assembled) product in the free-trade zone to have a minimum degree of local value added would reduce the discrepancy between gross exports and exports in terms of value added.

1. (15 points)ConsiderFigure 1. Whatdoesit show? Explain as precisely as possiblewhatismeasured on each axis, what a point represents, whatdoesitmean if a point is on, below or above the 45-degree line. Select tworemarkable country pairs illustrating the interest of the approach and explain the economic intuition behindit. Whydoesitmatter for «economicdiplomacy»? Suppose youwere a Chinesediplomat at the WTO. Whatwouldyou use this figurefor? Whatwouldyousuggest to the US administration on its basis ? Do youexpectit to apply to the EU as well? Why or why not?

Figure 1

Note: as one moves away from the origin, for country pairs involving the U.S., the U.S. trade deficit becomes larger.

Source: Koopman et al. 2011

Figure 1 shows the gross balance of trade, i.e. the conventional one thatismeasured in tradestats, on the horizontal axis. On the vertical axis, it shows the balance of value-addedtrade. A point is a pair of countries.
A country pair on the 45-degree line is one for which the bilateraltrade balance isequal, whethermeasured in gross or value-addedflows. There isonly one in that case, US vs. ROW (no particularreason for that). A country pair below the 45 degree line is one for which the US deficitishigherwhenmeasured in grossflows; one above the 45 degree line is one where the US deficitishigherwhenmeasured in value added.
Whatisremarkableis the position of the US-China and US-Japantrade balances whichstraddle the 45-degree line. With China, the US runs a largerdeficit in grossflows (below the line). WithJapan, the US runs a largerdeficit in value added (above the line). This isbecause China assembles electronicsproductsbefore export to the US; most of the expensive components of thoseproducts are made in Japan. Therefore the US deficitwith China picks up the full value of thoseproducts, whereasonly a small part of themis made in China. The opposite istruewithJapan: components exported by Japan to China for assembly and re-export to the US are not picked up in the US tradedeficitwithJapan.
This observation isvery important and matters a lot for economicdiplomacy. The Clinton, Bush and Obama administrations have all accused the Chinesegovernment of deliberatelyundervaluing the Chinesecurrency in order to get an unfaircompetitiveadvantagewhich translates intoheavy US tradedeficits and «steals jobs» from the US.
A Chinesetradediplomatwouldbewelladvised to use Figure 1 to argue that the US is «barking up the wrongtree» as the real competitivenessproblemiswithJapan, not China. China does not competewith the US, becauseitdoeslabor-intensive assembly in which the US has no comparative advantageanyway. It isJapanthatcompeteswith the US in the production of capital-intensive and skill-intensive components. The US should talk about exchange rates withJapan and leave China alone (sosays the diplomat).
The exact same argument applies to the EU which imports the sameelectronicsassembled in China out of Japanese components.

Question 2(multiplechoice)

Each multiple-choice question isworth5 points.Pleaserespond on the grid on page 7.

1. ConsiderFigure 2 in whicheach point represents a country. The vertical axis measuresgrowth in per-capita GDP between 1980 and 1990, and the horizontal axis measures GDP per capita in 1980, in logs. The scatter plot shows

1. Conditional convergence
2. Absolute divergence if the slope of the line issignificantlydifferentfromzero
3. Heteroskedasticity
4. The scatter plot cannotbeinterpretedbecauseitincludesonly OECD countries
5. None of the above.

Figure 2

Source: World Bank, World Developmentindicators

2. In Figure 3, the horizontal axis measuresaveragegrowth in GDP per capita during a set of growthaccelerations, while the vertical axis measuresaverageincomegrowth for the bottom quintile during the samegrowthaccelerations. The figure shows that

1. Growthisregressiveexceptduringaccelerations;
2. The figure cannotdemonstrateanything about growth and povertybecause of selectionbias;
3. The figure cannotdemonstrateanything about growth and povertyunlesswealso have the evolution over time of the Gini coefficient;
4. Nothing; wecannotsayanything about the relationshipbetweengrowth and povertywithout running a panel regressionwith country and mostimportantlyyearfixedeffects;
5. Growthaccelerations are neitherregressivenor progressive providedthat the slope of the regression line is not significantlydifferentfrom one;

Figure 3

Source: Dollar et Kray (2005)

3. Suppose that in a randomized-control trial, insecticide-treated nets (ITNs) are sold at 40 shillings, (a fraction of theirproduction cost) to a treatment group of pregnantwomen in Kenya. Suppose alsothatexposure to malaria ismeasured by the rate of hemoglobin in blood (alower rate meansmoreexposure to the disease). The cumulative distribution of hemoglobin rates isshown in Figure 4,with dots for the treatment group and emptycircles for a control group representing the population at large.

1. Under the selectionhypothesis, the curve for the treatment group shouldlie to the right of the curve for the control group.
2. Under the selectionhypothesis, the curve for the treatmentgroup should lie to the left of the curve for the control group.
3. Under the moral hazardhypothesis, the twocurvesshouldberoughlyidentical(i.e. thereshould not be a statisticaldifferencebetween the two)
4. The figure has nothing to do witheitherselection or moral hazard; itis about the priceelasticity of demand for ITNs
5. None of the above

Figure 4

Source: Cohen et Dupas (2010)

4. Malaria wassuccessfullyeradicated in Indiabetween 1950 and 1965 roughly. Consider the followingapproach to estimatewhether malaria creates a viciouscircle of lowachievement, poverty and disease. Let be individual i’s educational achievement at t, be his/her residence district, a dummy variable equal to one if the individual was born after the period of eradication (completed in 1965), andthe initial incidence of malaria in districtd.We estimate:

(1)

whereis a vector of individual characteristics observed at timet.Suppose that is positive and significant. This suggests that

1. Malaria eradication raises future educational achievement
2. Educational achievement is higher for individuals born after eradication in districts with heavy initial infestation
3. There is convergence in educational achievements across districts in India
4. Nothing as long as the equation does not include time and district fixed effects
5. None of the above

5. Suppose that we want to estimate the effect of a program of technical assistance to managers of small and medium enterprises (SMEs) in a given country on the performance of their company. Suppose also that (i) estimated treatment effects are not significant; (ii) regressing the performance of control firms on encounters between their managers and those of treated firms returns a positive and significant effect at the 1% level, this effect being robust. The results suggest that

1. The program is ineffective and should be terminated, because only (i) matters
2. The program should be continued with the same treatment group, but on a full-cost recovery basis (service for fee)
3. There is absolutely no way the program can be evaluated if the treatment effect «leaks»;only a randomized-control trial would generate hard evidence on treatment effects in the presence of such leaks
4. The program should be extended to all firms but should not be subsidized
5. The results are encouraging and should be confirmed using a better control group; at least the program passes the market-failure test (generates a positive externality)

6. The paper by Nunn and Qiang (2011) suggests that

1. US food aid may encourage conflicts in beneficiary countries
2. US food aid goes primarily to fragile and failed states (like Congo, the Central African Republic, or Afghanistan)
3. US development aid is targeted to countries whose vote patterns at the UN are aligned with US interests
4. US development aid does not go to corrupt countries because of the adoption by the US of the Foreign Corrupt Practices Act
5. None of the above

7. Suppose that we want to estimate the effect of a training program with the following difference-in-differences regression

(2)

whereiis an individual,is an individualfixed effect,ta year, a dummy variable equal to one when the individual is «treated» by the educational program (during his/her treatment period) and zerootherwise, and a set of relevant characteristics of invididualithat could possibly correlate with performance. Suppose that enrolment in the program and is voluntary. Propensity-score matching

1. Reduces selectionbiasin (2)by assigningweights to control-group individuals on the basis of how similar they are with treated invididuals, using as a summary statistics their predicted probability of treatment on the basis of a first-stage regression where treatment status (treated or not) is the dependent variable
2. Eliminates selection bias by matching each treated individual with another treated individual, using as a summary statistic theirpredicted probability of treatment on the basis of a first-stage regression
3. Is a technique that consists of using an instrumental variable for treatment status, with the instrument satisfying two basic requirements (exclusion restriction and correlation with the endogenous regressor)
4. Is an econometric technique that consists of estimating the odds of treatment and using them as an instrument
5. Has nothing to do with the estimation of treatment effects, which can be estimated only using randomization.

8. Consider the effect of social capital, measured by the degree of trust between people and by the degree of civic behaviour, both measured in surveys, on growth. Resultsfrom a Barro growthregression are shown in Table 1.When instrumented, trust is instrumented by ethnolinguistic fragmentation. The results suggest that

1. Civic behaviour and trust correlate with growth but trust weakens convergence
2. Trust behaviour correlates with growth but not when instrumented, in which case the effect is no longer significant at 5%
3. Trust correlates with growth at the 5% level or above in all specifications and reinforces convergence
4. Trust and civic behaviour apparently correlatewith contemporaneous growth but the results are suspicious since they are incompatible with conditional convergence
5. Instrumental variable estimation makes no sense in this case because ethnolinguistic fragmentation is a subjective measure that is itself influenced by culture etc.

Table 1

Note: standard errors in parentheses

Source: Knack and Keefer (1997)

9. The World Trade Organization’s agreements

1. Explicitly prohibit the use of non-tariff barriers through Article XI of the GATT
2. Permit the imposition of restrictive measures on products in the case where conditions of production are deemed unacceptable; in that case the measures must be justified (Article XX of the GATT)
3. Prohibit the use of anti-dumping regulations unless those use the «zeroing» calculation method, in which case they must be cleared with the WTO Secretariat (Article 6 of the anti-dumping agreement)
4. Allow the imposition of restrictive measures on imports provided that those measures are based on scientific evidence (Article 5.2 of the TBT agreement)
5. Make it illegal to ask questions about the WTO in bachelor exams, especially at the end when everybody is getting tired and just wants to get off the damn overheated room (Article I of the Unfair Measures agreement).

ANSWER GRID

(PLEASE CIRCLE THE CORRECT ANSWER):

Example of correct formatting (b being the right answer):

0. / a / b / c / d / e
0. / a / b / c / d / e

Example of wrong formatting(b being the right answer):

1. / a / b / c / d / e
1. / a / b / c / d / e

Answers:

1. / a / b / c / d / e
2. / a / b / c / d / e
3. / a / b / c / d / e
4. / a / b / c / d / e
5. / a / b / c / d / e
6. / a / b / c / d / e
7. / a / b / c / d / e
8. / a / b / c / d / e
9. / a / b / c / d / e

1