Week 10: Political Economy

A, B; Adding in country fixed effects to a regression is equivalent to subtracting average growth for a specific country from the left hand side of the regression and subtracting average temperature growth for a specific country from the right hand side of the regression. We can think of this as demeaning the data, which allows us to isolate the impact of within-country variation in temperature over time.

Adding in country fixed effects _________________________ (Check all that apply) a. Is equivalent to subtracting country averages from both sides of the regression b. Allows us to isolate the impact of within-country variation in temperature on growth c. Allows us to isolate the impact of across-country variation in temperature on growth

B; Looking across countries, a pattern emerges where high temperatures tend to be associated with low growth. For example, temperatures are highest in Niger (27 or 28 depending on the year) while growth is low (0.1 for both years included in the data); in comparison temperatures in the USA are relatively low (12 and 14 for each of the years) and growth is relatively high (3 and 3.2 for each of the years).

Using the sample data for temperature over time for various countries, what is overall relationship between temperature and growth, looking across countries? a. High temperatures seem to be correlated with high growth b. High temperatures seem to be correlated with low growth c. No clear pattern or relationship

All of the above are possible reasons that could explain why warmer countries tend to be poor. There are of course many exceptions to this rule. For example, Singapore is very warm and rich, while Afghanistan is cold but poor.

What are some of possible reasons that warmer countries tend to be poor? (Check all that apply) a. Prevalence of disease in warm climates b. Warm climates might limit agricultural productivity c. Warm temperatures may impact productivity

All; All of the above found to be lasting impacts of the mining Mita. One interpretation is that haciendas chose to locate outside the Mita boundaries, to avoid being required to send a fraction of their men to work in the mines. These haciendas made investments in public goods such as constructing roads and schools. These developments and investments were concentrated outside the Mita boundary, rather than within the Mita boundary, and the resulting differences have persisted to the present.

According to the Mita results, what were some of the lasting impacts of the Mita in the areas where a fraction of men were required to be sent to work in the mines? (Check all that apply) a. Fewer haciendas established b. Lowered investment in human capital c. Lowered investment in education d. Less investment in infrastructure such as roads e. Lower income

A; The point estimates using both measures are similar in size, but they are statistically significant only using distance to Potosi, as in the second set of regressions. The difference is probably driven by the fact that distance to Mita is a more economically meaningful variable, which captures a larger share of the residual variance of the outcome variable, thus increasing the precision of the estimates.

Comparing the results from the first set of regressions (using latitude and longitude as in questions 3 to 9) to the results from the second set of regressions (using distance to Potosi, as in questions 10 to 14), you can see that the coefficients estimates are relatively similar, but the level of significance differs between the two sets of regressions. 75km is significant at 1%, 100km is not. Which of the following might explain why that is the case? A. Distance to Potosi captures more of the residual variance of the outcome variable B. Latitude and longitude are unreliable measures C. There is not enough variation in latitude and longitude in the sample, so the impact of latitude and longitude on any outcomes is imprecisely estimated D. Distance to Potosi is likely mis-measured

B; Outcomes are compared across the border specifically along the portion of the border where those areas just inside and just outside have similar elevation. In other areas, the border coincides with a change in elevation, and it could be difficult to disentangle the effect of the difference in elevation and other related climatic factor from the difference between being inside or outside the Mita boundary.

Dell (2010) chooses to compare outcomes across the Mita border specifically along one area of the border, instead of along the whole border. Why? a. More accurate data is available for those areas b. To avoid areas that have differences in elevation just along the border c. It would not be feasible to look at the whole border, and these areas were randomly selected d. Because these areas were in between the largest mines in Potosi and Huancavelica

C; Dell cannot simply compare districts that formerly had the Mita system in place to those that did not. This problem arises because these regions might be different for several reasons other than the legacy of the Mita institution itself. For instance, they are probably more rural and better suited to mining, so they lack exposure to development driven by manufacturing or high-tech services. In addition, Dell notes the terrain is significantly different, as is their access to infrastructure. Most of these factors would lead us to expect Mita regions to be poorer even the institutions were exactly the same. The key identifying assumption is that the only difference between the two sides of the border is the variable of interest, in this case poor colonial institutions. Thus, even if there are average differences between places with and without Mita institutions, those differences are likely to be small right around

Dell uses a regression discontinuity design to compare outcomes between people who live in former Mita districts with those who live in districts without the Mita. This design exploits the fact that: A. Outcomes just inside and just outside the Mita border turned out to be very different, regardless of any differences in underlying characteristics at the start B. People very close to the border were least likely to have migrated from a Mita to a non-Mita district C. Conditions just inside and just outside the Mita border would likely have been very similar in the absence of the mining Mita D. People very close to the border were most likely to have migrated from a Mita to a non-Mita district

False; As Professor Olken mentions, national borders in Africa often do not line up with ethnic borders. The European colonial powers drew up national borders based largely on sphere of influence rather than any with any consideration for what boundaries would lend themselves to the most successful nation states.

European colonial powers typically drew up national boundaries in Africa based on the distribution of ethnic groups and ethnic boundaries. a. True b. False

B; The treatment effect is represented as the "jump" in outcomes along the outcome variable that is modeled to have been due solely to the treatment or imposition of the cutoff at c. The key assumption is that outcome variable would have carried along in a predictable way (similarly just before and after the cutoff) and that the treatment introduces this "jump," the treatment effect. This is represented by the distance from B to C.

In practice, where by definition, we do not observe the counterfactual outcome in the absence of the treatment, which represents the treatment effect? a. The distance from A to B b. The distance from B to C c. The difference between the slope of the line to the left and the slope of the line to the right of c d. The ratio of the slope of the line to the left the slope of the line to the right of c

B; In a regression discontinuity analysis framework, the Xi variables are the independent variables, or the underlying characteristics based on which the criteria or cutoff is determined. In this case, the Xi variables are distance to the Mita border or latitude and longitude, and treatment is determined by whether you are inside or outside the border.

In the Mita example, what are the Xi variables? a. A range of outcomes, including road construction, school construction, income today, etc. b. Distance to the border of the Mita (or latitude and longitude), indicator variable for inside or outside the Mita c. Number of years for which the Mita was in place

False This is false. It is the case that the result that institutions impact economic growth persists, even when controlling for temperature. However this does not preclude temperature also being a determinant of economic outcomes and economic growth within a country.

In the framework discussed in class, Professor Olken sets up a simplified model of economic outcomes as a function of institutions and temperature. From the evidence discussed previously, we know that institutions matter, even when controlling for temperature. True or false: This provides evidence against the argument that high temperatures cause lower economic growth. a. True b. False

B; In the framework outlined in class, the relationship of interest is to examine the impact of levels of trust in society on income. The reduced form captures the relationship between the slave trade and income today, where trust might be the missing endogenous variable. The first stage captures the effect of the slave trade on trust.

In the instrumental variables framework outlined in class, the first stage captures the ________________ and the reduced form captures the _____________, where ________ is the endogenous variable. a. effect of slave trade on levels of trust ; effect of slave trade on income; slave trade b. effect of slave trade on levels of trust ; effect of slave trade on income; trust c. effect of slave trade on income; effect of slave trade on levels of trust ; slave trade d. effect of slave trade on income; effect of slave trade on levels of trust ; income

b; β represents the coefficient estimate of temperature on growth for rich countries. γ represents the additional impact of a one degree increase in temperature on growth for those countries classified as a poor country. So, the total impact of a one-degree increase in temperature on growth would be β + γ. For a rich country, the dummy variable for poor would equal 0 (POOR=0). Then, the effect of a one degree temperature increase (TEMP=1) on growth would be β(1)+ γ(1)(0), which equals β. For poor countries, POOR=1, and the effect of a one degree temperature increase on growth would be β(1) + γ(1)(1).

In the regression above, what is the total effect of a one-degree hotter year in a rich country is ____________ and the total effect of one-degree hotter year in a poor country is __________? a. β−γ ; β+γ b. β ; β+γ c. β+γ ; γ d. β+γ+α; γ+α

B; Explanation: Michalopoulos and Papaioannou combine the light intensity data with data on pre-colonial ethnic boundaries and institutions, and the fact that national borders were drawn artificially, to look at how much national institutions matter once ethnicity is controlled for.

In their paper looking at the impact of ethnic institutions on development, Michalopoulos and Papaioannou use which of the following to see how economic activity changes across national borders within ethnic boundaries? a. Data on cross-border trade flows b. Data on light intensity at night c. Data on flows of goods and services between pre- and post-colonial periods d. Settler mortality data

B; Acemoglu, Johnson, and Robinson hypothesize that colonizers were more likely to settle in places where they would want to live or were hospitable in some way, and would set up extractive institutions in places where they were only interested in extracting wealth. They set up a first stage which is to look at the impact of settler mortality (the idea being that colonizers would not want to live in places with high settler mortality) on the existence of extractive institutions (the idea being that high settler mortality might be associated with extractive institutions in the past, that persist as poor institutions in the present).

In their research, Acemoglu, Johnson, and Robinson examine the impact of ____________ on ___________, which in this analysis is the __________________. a. Settler mortality ; existence of extractive institutions ; Reduced form relationship b. Settler mortality ; existence of extractive institutions ; First stage c. Settler mortality ; per capita income today ; First stage d. existence of extractive institutions; per capita income today; First stage

C; The exclusion restriction requires that the only way settler mortality impacts GDP, is through institutions. If settler mortality impacted some third variable, which in turn also impacted GDP, then settler mortality would not be a strong instrument for institutional quality. As Professor Olken discusses in class, the exclusion restriction cannot be directly tested, but has to be made as an assumption (and can be supported by a variety of indirect tests).

In this analysis, which correctly describes the exclusion restriction? a. Settler mortality is very highly correlated with current expropriation risk b. Settle mortality is not correlated with current expropriation risk but is correlated with current GDP c. The only way that settler mortality impacts GDP is through its effect on institutions d. The only way that current expropriation risk impacts GDP is through settler mortality

C; In the first-stage, Acemoglu, Johnson, and Robinson examine the relationship between historical settler mortality and current expropriation risk. Current expropriation risk acts as a measure for the quality or strength of institutions today.

In this example, expropriation risk is used as ________________________. a. A direct measure of current institutional quality b. A direct measure of historical institutional quality c. A proxy for the current institutional quality d. A proxy for the historical institutional quality

C; Michalopoulos and Papaioannou do not explicitly control for national language or geographical factors. The level of light intensity is the dependent variable in their analysis i.e. it is the proxy for development that they use, not an explanatory factor. They control for ethnicity by including ethnicity fixed effects in their regressions and when they do so, they find that the impact of national institutions - such as the rule of law and degree of corruption - is not significant any more.

Michalopoulos and Papaioannou find that national institutions affect contemporary development, but this effect disappears when: a. The national language is controlled for b. Geographical factors are controlled for c. Ethnicity is controlled for d. The level of light intensity is controlled for

False; False. In order for RDD to be a valid analytical approach, it must be that the underlying characteristics or counterfactual outcome are continuous, but that there is a jump in treatment or cutoff variable. In other words, the limit from below must be equal to the limit from above the cutoff for the underlying characteristic or counterfactual outcome, but the limit from below must be unequal to the limit from above the cutoff for the treatment or cutoff variable. In mathematical terms, it should be that: 𝑙𝑖𝑚𝑥↑𝑐𝑌𝑖(0)|(𝑋𝑖=𝑥)=𝑙𝑖𝑚𝑥↓𝑐𝑌𝑖(0)|(𝑋𝑖=𝑥) 𝑙𝑖𝑚𝑥↑𝑐𝑇𝑖|(𝑋𝑖=𝑥)≠𝑙𝑖𝑚𝑥↓𝑐𝑇𝑖|(𝑋𝑖=𝑥)

Let Yi(0) = the counterfactual outcome, Ti = the treatment variable, and x = an underlying characteristic, and c = the cutoff for treatment. True or false: In order for RDD to be valid analytical approach, it must be the case that: limx↑cYi(0)|(Xi=x)=limx↓cYi(0)|(Xi=x) limx↑cTi|(Xi=x)=limx↓cTi|(Xi=x) a. True b. False

All; Broadly speaking, poor countries do not seem to be randomly distributed throughout the globe. Of course there are many exceptions, but countries along the equator and with warm climates tend to be poor. Former colonies and countries in sub-Saharan Africa tend to be poor. This unit will delve into some of the micro-determinants and macro-determinants of development.

Looking at relative income across countries, which of the following seems to be true about poor countries? (Check all that apply) a. Former colonies tend to be poor b. Countries along the equator tend to be poor' c. Warm countries tend to be poor d. Countries in sub-Saharan Africa tend to be poor

B, C; B and C are both possible ways in which the slave trade could have impacted the development of institutions, which could have lasting impacts into the present day. One of the main hypothesis that will be tested in this research is that the slave trade relied on people cooperating with slave traders to sell or transport slaves. This could have led to a culture of mistrust which could have impacted and weakened institutions. To the extent that those institutions have persisted into the present, this would affect outcomes today. Similarly, many countries that were involved in the slave trade lost a significant portion of their population. Changes in the population composition could have had meaningful and lasting effects. A is also likely true, but this is not a reason why the slave trade would matter.

Nunn (2008) looks at some of the long-lasting impacts of the slave trade in Africa and how that impacts current institutions and outcomes. In what ways might we expect that the slave trade could impact outcomes? (Check all that apply) a. Countries that were very involved in the slave trade tend to be those with high temperatures, and as we saw from before, these countries tend to have lower growth b. There is less trust among people where the slave trade was historically very predominant, this impacted the development of institutions into the present c. Countries where the slave trade was prevalent lost many young and healthy people, and these population changes could have lasting impacts into the future

C, not A. The first stage captures the relationship between distance to a port and the volume of slave trade. A strong first stage would find a negative relationship between distance to ports and volume of slave trade, where those countries further from a port trader fewer slaves than those closer to a port. As shown in the below table, Nunn (2008) finds a strong first stage, where the effect of distance to a port on slave exports is negative and significant.

Nunn (2008) uses distance to major slave trading ports as an instrument to estimate the impact of the intensity of slave trade on income today. What does Nunn look for to provide evidence for or against the strength of the first stage? a. A negative and significant relationship between distance to slave ports and income today b. No significant relationship between distance to slave ports and income today c. A negative and significant relationship between distance to slave ports and volume of slave trade d. No significant relationship between distance to slave ports and volume of slave trade

D; The estimate for β + γ, or the impact of a one-degree hotter year on economic growth for a poor country can be calculated by adding the separate estimates for β (0.261) and γ (-1.655), or -1.394. While the calculation in C is correct, the standard errors cannot be similarly added, and we cannot determine whether or not the estimate is significant. However, the combined impact of β + γ is also reported at the bottom of column 2 as -1.394, and the standard errors are correctly computed and presented to indicate that the estimate is statistically significant, as in D.

Refer to the results presented in class, and in the above table for the regression results presented in column 2. Which of the following is correct about the estimate for β + γ? a. 0.261 and insignificant b. -1.655 and significant c. 0.261-1.655 = -1.394 and insignificant d. -1.394 and significant

B and C only; Regression discontinuity design rests on the assumption that there are certain programs or policies where a strict eligibility criterion applies, but that the underlying characteristics or outcome in the absence of eligibility criteria are probably continuous. In the example given in class, people just below the age of 21 are probably very similar to people just above the age of 21 along meaningful characteristics such as maturity, ability to drink responsibly, physical development, etc. However, someone just below 21 is unable to legally drink US, while someone above age 21 is able to legally drink in the US. We can think of people below 21 being very close to identical to people above 21, aside from the important distinction of being legally unable to drink alcohol in the US.

Regression discontinuity design rests on the assumption that: (Check all that apply) a. Underlying counterfactual outcomes lend themselves to a natural or clear cutoff b. Underlying counterfactual outcomes are continuous, while corresponding eligibility cutoffs are discontinuous c. People just above and just below a cutoff are very similar d. Eligibility cutoffs are arbitrarily chosen

A, B; As Professor Olken discusses in lecture, there seems to be a strong relationship between temperature and economic performance. However, it is difficult to tease out the impact of temperature from other characteristics of a country that are fixed, and that might also impact economic growth. For example, it could be that warmer countries were more likely to have been colonized in the past, and that the legacy of colonization has impacted growth in the present day (as in A). Similarly, there could be other fixed characteristics that are correlated with temperature and also impact growth (as in B).

Temperature seems to be correlated with low economic growth and performance. What are some reasons we cannot immediately conclude this is a causal relationship? (Check all that apply) a. There are many other factors correlated with temperature that we might expect would impact economic growth (colonial history, etc.) b. Temperature could be correlated with other fixed characteristics that also impact growth (longitude, elevation, etc.) c. There are no theoretical reasons to believe temperature would impact economic growth and performance

A; If light intensity is understood as a proxy for development, with greater light intensity meaning a higher level of development, and if national institutions did indeed matter for development, then we would expect to see more light intensity in the blue area of the map (Ghana) than in the yellow area (Togo.) That is, even within the same intensity, these differences in development would arise owing to national institutions and their influence on development.

The image below depicts the distribution of the Ewe ethnic group in West Africa. If we believed that Togo has worse national institutions than Ghana and that this matters for development, then Michalopoulos and Papaioannou hypothesize that we should see: a. More light intensity in the blue versus yellow area b. More light intensity in the yellow versus blue area c. Less light intensity in the yellow area versus the white area d. Similar amounts of light intensity in the yellow and blue areas

True. Acemoglu, Johnson, and Robinson's paper hypothesizes that colonizers set up different institutions in colonies they planned to settle compared to colonies from whom they planned to extract resources or otherwise exploit. In those colonies that were established as a way to extract resources and wealth, settlers set up "extractive institutions," which have had a lasting and negative impact on colonies.

True or False: Acemoglu, Johnson, and Robinson hypothesize that colonizers set up very different institutions in colonies where they planned to settle compared to those from whom they simply wanted to extract resources. a. True b. False

True; True. Using settler mortality as in instrument for the quality of institutions, Acemoglu, Johnson, and Robinson (2001) find evidence that past institutions persist into the present, and that these institutions matter in that they impact income today.

True or false: Acemoglu, Johnson, and Robinson find that institutions do impact economic outcomes in a country. a. True b. False

This is false. If it is determined that countries do not have worse outcomes during hotter years, it could be simply that long-run temperatures matter more for determining economic output than temperature or variation in any single year.

True or false: If it is determined that countries do not have worse outcomes during hotter years, this would provide strong proof that temperature does not impact growth or economic outcomes in developing countries. a. True b. False

True; In practice, the researchers must consider a few factors to which the analysis might be sensitive. As discussed in lecture, the analysis could be sensitive to how close in to the cutoff researchers look. For example, it could be that people who are 1 unit below the cutoff are very similar to people that are 1 unit above the cutoff, but the results may look different if we include people who are 10 units above or below the cutoff. Often in RDD analysis, researchers will repeat their analysis on different samples, each time varying distance from the cutoff.

True or false: RDD analysis may be sensitive to how close or far from the cutoff c that researchers consider. a. True b. False

False; The exclusion restriction requires that the only way that distance to a port impacts income today is through the slave trade, which cannot be directly tested. As we've discussed in other units in the class, the exclusion restriction cannot be tested directly in the data.

True or false: The exclusion restriction can be directly tested to determine the validity of distance to a slave trading port as an instrument a. True b. False

True; True. Dell (2010) runs the tests described in the previous lecture segment for areas 100, 75, 50, and 25 kilometers away from the border, and finds that these areas become more and more similar in terms of elevation, slope, and tribute paid to the Spanish crown as distance to the border decreases. This provides some support that the border is a valid discontinuity across which to compare outcomes.

True or false: The results demonstrate that as you get closer and closer to the border of the Mita, the areas become more and more similar a. True b. False

All of the above are ways the mining Mita might have impacted areas affected by the Mita. Knowing they were likely to have to work in a mine might have shortened the horizon over which adult males made investment decisions, and may have specifically discouraged them from making human capital investments (as in A and B). Sending a large portion of the adult male population to work in the Mita would have drastically changed the composition of the population (as in C). It could also be that certain types of people would have left the affected areas, which would have further changed the composition of the population.

What are some of the reasons that the mining Mita might have affected outcomes during the time the system was in place? (Check all that apply) a. Shortened time horizon for making decisions b. Lowered investment in human capital c. Population composition changes d. Outmigration from places affected by the system

B is correct. It could be that certain decisions related to the development of certain institutions, investment in infrastructure, roads, etc. were taken knowing which areas were within and outside the Mita boundaries. To the extend that some of these decisions have lasting impact, areas within historical Mita boundaries might have different outcomes compared to areas outside historical Mita boundaries, even today. This forms the key argument of Dell's 2010 paper.

What are some reasons the mining Mita might have affected outcomes even today, 200 years later? a. Some communities have chosen to maintain similar systems into the present b. Decisions about haciendas, roads, other infrastructure were made taking into consideration the existence of the Mita c. Communities that were within the Mita boundary were distinctly different than those outside the boundary to begin with, these differences have persisted to the present day

A; If we look at the relationship between temperature and growth within a country, it seems that high temperatures seem to be somewhat correlated with high growth. Even when a clear pattern or relationship between countries seems to indicate that high temperatures are correlated with low growth, there could be a different pattern that emerges looking within countries over time.

What is the overall relationship between temperature and growth, looking only within countries? a. High temperatures seem to be somewhat correlated with high growth b. High temperatures seem to be somewhat correlated with low growth c. No clear pattern or relationship

A, D; While one may question the validity of light intensity as a proxy for development, this is not one of the issues cited by Professor Olken. As he points out, ethnic groups in Africa typically straddle multiple nation states since the European colonial powers did not draw up national boundaries based on the distribution of ethnic groups. As he does note though, if borders are porous, people may travel across national boundaries and thus take the effect of their local institutions to other places. This would cause us to underestimate the impact of national institutions within ethnic groups. Another issue that he raises is that the effect of national institutions may just not radiate out to border areas and be largely concentrated in key cities; thus, development along the borders may not be a valid outcome for thinking about the impact of institutions on development.

Which of the following is raised by Professor Olken as possible issues with the Michalopoulos and Papaioannou paper? (Check all that apply) a. Porous borders might mean that ethnic groups travel across national boundaries and the effect of national institutions thus leaks onto neighboring countries b. Light intensity is a bad proxy for development c. Ethnic groups in Africa are generally contained within single nation states d. National institutions may matter in the main cities, but not in the border areas that are analyzed by Michalopoulos and Papaioannou

B; The running variable is a measure of location, either latitude longitude or distance to the Mita.

Which of the following is the "running" variable? (Hint: In general with Regression Discontinuity Design, the running variable refers to the underlying characteristic that is assumed to be continuous; at some cutoff or threshold value of this variable, as defined by policy, we expect a discontinuity in outcomes to arise.) A. A measure of income today B. A measure of location, either latitude/longitude or distance to the Mita C. A measure of current institutional quality D. A dummy variable for in or not in a Mita district

D; D is correct. In order to demonstrate or provide support that the discontinuity is valid, we would want to look for factors that are NOT related to the outcomes of interest today - that is, we would want to look for factors that we expect would have been unaffected by the existence of the Mita - and test whether these are significantly different based on distance to the border of the Mita. If such tests demonstrate that areas do not vary along these other factors based on distance to the border, then that would support that these areas were very similar to begin with, and that we can compare outcomes and attribute any differences to the introduction of the Mita.

Which of the following would support that using distance to the border as an independent variable in this analysis is valid? a. That outcomes of interest (education today, income today, etc.) ARE significantly different based on distance to the border b. That outcomes of interest (education today, income today, etc.) ARE NOT significantly different based on distance to the border c. That other factors (unrelated to income today, education rates today, or other outcomes of interest) ARE significantly different based on distance to the border d. That other factors (unrelated to income today, education rates today, or other outcomes of interest) ARE NOT significantly different based on distance to the border

A, C, and D are all possible negative outcomes of colonization that might have persisted even beyond independence. Colonizers might have set up boundaries in arbitrary and even harmful ways, and these boundaries have persisted into the present. It could also be that the end of colonization led to political unrest or instability, which has lasted into the present or otherwise disrupted economic development. One hypothesis we will discuss in this class is that some colonies might have been designed, not as a place for colonists to settle, but as a way to extract wealth and resources. In these cases, institutions might have been set up in such a way to help extract wealth and resources, which may have had lasting negative impacts on the former colony. B is incorrect, since it states that differences in outcomes for non-colonies and colonies are due to climate-related factors, and not due to colonization.

Why might former colonization have lasting impacts to the present day? (Check all that apply) a. Boundaries may have been drawn in detrimental ways, and these boundaries persist to today b. Warm countries along the equator were most likely to be colonized, these countries would have had worse economic outcomes anyways c. Colonizers set up institutions that were designed to extract wealth and resources, but not to invest in the long-term health of the colony's economy, infrastructure, etc. d. Political instability following independence

C, D; C and D are both correct. As discussed in the lecture, there could be factors common to certain regions within a given year (aside than temperature) that impact growth. If these are uncorrelated with temperature, leaving these factors out would reduce precision, as in C. However, if these factors are correlated with temperature, leaving them out of the regression would lead to biased estimates, as in D.

g_crt = α_c + γ_rt + βTEMP_crt + γTEMP_crt∗POOR_c + ϵ_crt In the regression above from Dell, Jones, and Olken (2011), why are continent-year dummies included (the γrt term)? (Check all that apply) a. This is a way of testing the robustness of the results of temperature on growth b. This is a way of controlling for differences over time, to isolate changes in growth over time only c. Certain years could be bad for reasons other than temperature in certain continents; leaving out the continent-year dummies would simply reduce precision even if those reasons are uncorrelated with temperature d. Certain years could be bad for reasons other than temperature in certain continents; leaving out the continent-year dummies would bias estimates if those reasons are correlated with temperature