Reading notes MT - Summary Development Economics PDF

Preview

Summary

WEEK 1 MT Beyond GDP? Welfare across Countries and Time by Jones C., Klenow P. Motivation To compare living standards of people in different countries taking into account welfare measures such as consumption, leisure, inequality, and mortality using the standard economics of expected utility. Specia...

Description

WEEK 1 MT Motivation

Special settings Theory

Empirical design

`Beyond GDP? Welfare across Countries and Time` by Jones C., Klenow P. To compare living standards of people in different countries taking into account welfare measures such as consumption, leisure, inequality, and mortality using the standard economics of expected utility. GDP is a flawed measure of economic welfare as many factors affect living standards within a country that are incorporated imperfectly, if at all, in GDP. Compare welfare across countries using a common specification for preferences. A utilitarian expected utility calculation giving equal weight to each person. The welfare metric could be an equivalent variation (by what proportion to adjust to equal welfare) and a compensating variation (by what factor to increase to raise welfare) producing different results especially for poor countries. The paper reports the equivalent variation. Calculate consumption-equivalent measure of welfare to see what proportion of consumption in one country, given its values of these factors, would deliver the same expected utility as values in another country and make an individual indifferent between two countries? It is assumed that utility from leisure takes a form that implies a constant Frisch elasticity of labour supply (holding the marginal utility of consumption fixed, the elasticity of labour supply with respect to the wage is constant). Various existing data sets (Penn World Table, UNU-WIDER World Income Inequality Database etc.) are used for data for a much wider set of countries. Evidence suggests that calculations using publicly-available multi-country datasets are potentially informative as they closely match the results derived from extensive data for the 13 aforementioned countries.

Data Detailed micro data from household surveys for 13 countries to provide a measure of welfare. Publicly available multi-country datasets to construct cruder welfare measures for 152 countries. The authors omitted morbidity (disease prevalence), the quality of the natural environment, crime, political freedom and intergenerational altruism. Key findings

Interpretation / Policy implications WEEK 1 MT Motivation

Special settings

GDP per person is an excellent indicator of welfare across the broad range of countries: the two measures have a correlation of 0.98. Nevertheless, for any given country, the difference between the two measures can be important. Across 13 countries, the median deviation is about 35 percent. Welfare is more dispersed than income. The way to reconcile these large deviations with the high correlation between welfare and income is that the “scales” are so different. Incomes vary by more than a factor of 64 in our sample, i.e., 6,300 percent, whereas the deviations are on the order of 25 to 50 percent. Average Western European living standards appear much closer to those in the United States when we take into account Europe’s longer life expectancy, additional leisure time, and lower levels of inequality. Many developing countries, including all eight of the non-European countries in our sample, are poorer than incomes suggest because of a combination of shorter lives, low consumption shares, and extreme inequality. Welfare growth averages 3.1 percent between the 1980s and mid- 2000s, versus income growth of 2.1 percent, across the seven countries for which we have household surveys during these periods. A boost from rising life expectancy of about 1 percentage point per year accounts for the difference. An important contributor to welfare growth is the increase in life expectancy (Table 3). Developing countries should focus on inequality issue. Through the findings of the paper, it seems that most developing countries ought to not simply focus on boosting incomes but boosting health and life expectancy directly. `Global Inequality of Opportunity: How much of our income is determined by where we live` by Branko Milanovic How much of a person`s income will be determined by a country of his residence, unrelated to individual effort or luck? Economic implications: where should the efforts of people in poor countries be directed: to work or to migrate? Less than 3% of the world`s population lives in countries where they were not born. More

Theory

Empirical design

Data Key findings

Interpretation / Policy implications WEEK 1 MT Motivation Special settings

Theory

Empirical design

than two-thirds of global inequality between individuals is due to national income differences. Income can be written as a function of country-specific circumstances, own specific circumstances whose effect also depend on country, person`s own effort and a random shock which can also be called luck. Each country is divided into 100 groups of equal size (percentiles) to compare the positions of, for example, the 23rd percentile of people in China with the 75th percentile in Nigeria and also allows to define income classes in the same way across countries. Incomes within all percentiles except the very highest one are extremely homogenous. Regressing the annual average household per capita income in $PPP on the country`s GDP per capita in PPP terms and inequality in income distribution. Both variables on the righthand side are strictly exogenous to an individual effort. Two approaches whether to take into account population sizes: the individual viewpoint (size is irrelevant) and the world as it is (accounting for size). To test the robustness replace GDP per capita with the average number of years of education of the population over the age of 15. To examine the location premium run the similar regression but with the person`s own income ventile (each ventile contains 5% of population, ranked from the poorest to the richest) held constant. For each ventile separately, regress ventile income on country`s GDP per capita and Gini coefficient. Use 2008 data, which comes from 118 countries’ household surveys representing 94% of the world’s population and 96% of world dollar income. Table 2: In the base case (regression 1), elasticity of own income with respect to country’s GDP per capita is 0.866. The Gini coefficient enters with a negative sign, indicating that living in a more unequal country on average reduces one’s income. Overall, these two circumstances explain two-thirds of the variability of individual percentile incomes across the world. The increase of a country’s average educational level by one additional year of schooling is associated with an increase of individual incomes of more than 30%. Table 4: if we take all people who are in a given ventile of their country’s income distributions (say, third or tenth ventile) some 90% of the variability of their incomes will be explained by GDP per capita and Gini coefficients of the countries where they live. While the locational premium holds for everyone, the premium is less for those in the lowest ventiles of income distribution. More than half of variability in income globally is explained by circumstances given at birth. The benefit of higher mean income is proportionately greater for the rich classes, so policy implications should target within country inequality. `Measuring Economic Growth from Outer Space` by Henderson J., Storeygard A., Weil D. To use changes in `night lights` as a measure of economic growth and a proxy for GDP growth In developing countries a much smaller fraction of economic activity is conducted within the formal sector, which makes GDP measurement problematic. Using lights, growth can be measured not only on a country level, but also for sub- and supranational regions as well as at a higher frequency. Identification strategy: consumption of nearly all goods in the evening requires lights: as income rises, so does light usage per person, in both consumption activities and many investment activities. Nevertheless, the percentage growth in income and lights is not expected to be the same. Combine the lights measure, which is in a different metric than income, with an income measure to improve estimates of true economic growth. Assume there is classical measurement error in GDP growth as recorded in national income accounts. The underlying assumption for the specification is that there is a simple constant elasticity relationship between total observable lights and total income. Experiment later with different functional forms and control for changes in dispersion of lights to prove that this underlying assumption is appropriate. Assume that measurement error in growth as depicted in the national income accounts is uncorrelated with the measurement error that occurs when the change in lights is used to measure growth.

Data

Key findings

Interpretation / Policy implications WEEK 3 MT Motivation

Special settings

Empirical design

Identification is from within-country relative variation in lights and income over time, relating growth and fluctuations in lights within countries to annual growth and fluctuations in measured income. Controlling for country fixed and year fixed effects. To compare growth between different areas divide up the continent into two or more zones based on a particular criterion then sum the digital number for all pixels in each zone and look at the log difference between the average for the first and last two years in the data. Compare this log change across zones allowing for both zone and time fixed effects. Night lights data collected from the satellites, which have been circling the Earth 14 times per day. Each satellite-year dataset is a grid reporting the intensity of lights for every 30 arc-second output pixel. Datasets currently exist for 30 satellite-years covering the years 1992 to 2008. Classify countries into different data quality groups using the World Bank grading, then isolate a group of very low-quality data countries. Table 2: col. 6: electricity consumption has essentially the same predictive power for GDP and the same elasticity as does lights. Table 5: from here use beta=1.15 as the point estimate of the structural elasticity of lights growth with respect to income growth, close to 1, so that the long-term rate of lights growth approximately equals the rate of true income growth. However, there are mechanisms that could push the elasticity both higher and lower than one: fixed costs associated with electricity distribution leading to a convex relationship (elasticity greater than one), diminution in the rate of increase of lights (more urbanization with people living above one another and economies of scale such as street lamps) producing the elasticity lower than one. For good data countries apply higher weight to the reported GDP growth rates (circa 0.85) but apply lower weight to bad data countries when estimating the average annual growth rate of true income. Table 6 repots measured income growth, predicted income growth from lights and the composite estimate of true income growth for bad data countries; there is divergence between official and estimated results. Compared to official measures, Table 6 shows that growth is more likely to be underestimated in the WDI (World Development Indicators) for countries with low measured income growth rates, and overestimated in the WDI for some countries showcasing very high growth rates Look at whether over the last 17 years coastal areas have grown faster than noncoastal areas; whether primate cities have grown faster than hinterlands; and whether malarial areas have had a better growth experience compared to nonmalarial areas. The answer to all these questions is no, which leaves for future research the question of why. Some aspects of growth are still being missed out even with the new measure. Measurement errors are probably being correlated despite the model`s assumptions. `The Lifecycle of Inventors` by Bell A., Chetty R., Jaravel X., Petkova N., Reenen J. Focus on the decision to become an inventor because inventors play a key role for growth and there are positive spillovers from innovation such that the social returns are greater than private returns; to see which types of policy tools are likely to be most effective in sparking innovation. We are able to follow potential inventors from their conditions at birth (parental income, gender and race), while growing up (neighborhood, school text scores, college attended) and finally in the labor market (their income profiles). We show that parental characteristics matter a lot: the rich, white and male are much more likely to grow up to be inventors than the poor, female and black. Early math test scores account for a third of the inventor gap for income, a tenth for the black-white difference and almost nothing for gender. Focusing on the differences by parental income, two elements prove to be key. First, the innovationrelevant human capital gap between rich and poor opens up as they go through school, and by the time they reach college parental income makes relatively little difference. Second, children who are most exposed to innovation through their parents, mentors or neighborhoods are much more likely to become inventors. Figure 1 shows (solid blue circles) the number of inventors per 10,000 individuals (left hand vertical axis) ordered by the parents’ percentile position in the national income distribution (x-axis). Measure the latter as average household income 1996-2000, which Chetty et al. (2014b) have shown to be a good proxy for permanent income.

Figure 2 shows the kernel density of third grade math test scores for rich and poor kids. We use a parental income split at the 80th percentile and label the “rich” those above this and the “poor” those below. Table A2: the dependent variable is whether the child grows up to be an inventor (scaled by 1,000). Measures of `exposure` to innovation: (i) whether the parent was an inventor; (ii) how innovative was the industry where a child’s parents worked and (iii) neighbourhood characteristics such as innovation in the childhood Commuting Zone (CZ). For all three dimensions we examine not just the overall level of innovation, but the type of innovation by exploiting the information on patent technology class. Table 3 puts the idea that the network of people in the firm and industry could influence what careers young people are interested in studying and pursuing in later life in a regression framework, where the data is constructed solely from all children whose parents were not inventors (in order to rule out the direct channels examined in the previous subsection). Table 4: Column (1) is the baseline regression where the dependent variable is the fraction of kids who lived in a commuting zone that grow up to be inventors and the key right-handside variable is the invention rate in the childhood commuting zone (as in Figure 11). Data Link data on all patents granted between 1996 and 2014 to federal income tax returns with information on over 1.2 million patent applicants or holders. Match inventors to taxpayers using inventors’ name, city, and state. Data from all individuals who went through the New York City (NYC) public school system between 1989 and 2009. Key findings

Children of low-income parents are much less likely to become inventors than their higherincome counterparts (as are minorities and women). Decompositions using third grade and older test scores indicate that this income-innovation gap can largely be accounted for by differences in human capital acquisition while children are growing up. Fig. 1: Children born to the richest 1% of parents had invention rates of 8.3 in every 10,000, which is an order of magnitude higher than the proportion of inventors born in the bottom half of the income distribution (0.85). The green triangles corresponding to the right-hand vertical axis show that the positive relationship between parental income and invention is just as strong for high-quality patents as it was for all patents. Fig. 2: The rich kids’ distribution is strongly shifted to the right as we would expect. Only 7% of poor kids are in the top decile of the math score distribution compared to 23% of rich kids. Table A2: Column (1) shows that third grade math scores are highly predictive of becoming an inventor - a standard deviation increase in math test score is associated with a (highly significant) increase of 0.85 in 1,000 chance of becoming an inventor. The results in following columns imply that early math ability is very informative for future inventor status whereas English performance is not. Panel B of Figure 3 shows that there is a positive relationship between early math test scores and the chances of becoming an inventor for both rich and poor children and again, the relationship is noisy until we get to the top decile of the math test score distribution. It is striking, however, that for children in this top decile, rich kids have a much higher invention rate than poor kids. Figure 5 illustrates the relative similarity of the distribution of math test scores across both genders. In stark contrast to income, the gender gap in innovation does not appear to be ability related. Figure 6 uses the NYC data (where we can observe race) to show that there are wide disparities in patenting races by minority status.18 The first blue bar shows that white children have an inventor rate of 1.6 in 1,000, which is more than three times the rate for black kids (0.5) and eight times the rate for Hispanics (0.2). It is amongst the most gifted at math that the differential invention rates by race become clearly visible. For third graders in the 10% of the math test score distribution, future inventor rates are over 8 for Asians, about 4 for Whites and about 1 for Blacks and Hispanics. As children get older, test scores account for more of the inventor-income gap. By 8th grade 53% of the gap is accounted for, compared to 30% in grade 3. On average an extra

Interpretation / Policy implications

WEEK 3 MT Motivation Special settings

Theory

Empirical design

4.4 percentage points of the gap is accounted for each year by test scores and the null hypothesis that there is no additional explanatory power of the later grades is rejected. Table A7 shows the sharp positive gradient between being born into a wealthy family and attending an innovative college. 1) Amongst those whose parents were inventors, the patent rate was 11.1 per 1,000. By contrast, if a child’s parent was not an inventor then the patent rate was only 1.2 per 1,000. On average, if an individual’s parent invented in a particular technology class, the child had over a 0.9 in a 1,000 chance of inventing in exactly the same class, whereas the chance of inventing in the next `closest` technology class was under 0.2, about a fifth as high. 2) Table 3: a strong positive and significant relationship between the proportion of workers in their father’s industry who were inventors (right hand side variable) and The dependent variable is the proportion of children who became inventors (within a father’s industry). 3) Table 4: children who grow up in innovation-intensive neighborhoods are more likely to become innovators themselves. According to column (1) of Table 4, increasing the fraction of inventors in the childhood commuting zone by a standard deviation (0.0002) is associated with a 30% increase in invention rates. `Exposure` to innovation in childhood has a strong association with the chances of growing up to be an inventor. The key findings from our labor market analysis are (i) there are substantial returns before the patenting itself; (ii) patenting returns appear very skewed: there is a small chance of a very high payoff (cf. Hall and Woodward, 2010, on entrepreneurs) and (iii) many returns are late in an inventor’s career (cf. Jones, 2009, 2010). Hence pay-offs are highly uncertain when individuals are making initial career choices. A lack of opportunity for those who would have had a comparative advantage in innovation but for their financial position leading to a loss of innovation and output due to a misallocation of talent, thus space for policy makers to remov...