Working Papers
Applying Synthetic Control Methods to Concurrent Measures of Nighttime Lights (with Yuansen Li)
Satellite measures of nighttime light (NTL) have proven useful for the study of economic developments in situations where on-the-ground data are scarce and/or of poor quality. The NTL data available prior to 2012 have known sources of non-random measurement error present in data captured by multiple satellites with relatively brief but overlapping lifespans. We show that a synthetic control method (SCM) that exploits data from overlapping satellites can construct credible counterfactual series of measured NTL. The procedure is even capable of constructing synthetic series that closely match NTL measurements taken by satellites that had not yet been launched during the SCM’s pre-treatment training period. We also identify jackknife+ methods as valid for producing predictive confidence intervals in this context. These analytical developments jointly enable evaluation of localized shocks that occurred as far back in time as the late 1990s. We offer two illustrative applications, each quantifying causal changes in NTL as the result of a shock. We estimate changes in NTL emitted by 1) the city of Banda Aceh, Indonesia as a result of the 2004 tsunami and subsequent recovery efforts, and 2) the capital city of Hugo Chavez’s home state of Barinas, Venezuela after his ascension to Venezuela’s Presidency in 1999.
Satellite measures of nighttime light (NTL) have proven useful for the study of economic developments in situations where on-the-ground data are scarce and/or of poor quality. The NTL data available prior to 2012 have known sources of non-random measurement error present in data captured by multiple satellites with relatively brief but overlapping lifespans. We show that a synthetic control method (SCM) that exploits data from overlapping satellites can construct credible counterfactual series of measured NTL. The procedure is even capable of constructing synthetic series that closely match NTL measurements taken by satellites that had not yet been launched during the SCM’s pre-treatment training period. We also identify jackknife+ methods as valid for producing predictive confidence intervals in this context. These analytical developments jointly enable evaluation of localized shocks that occurred as far back in time as the late 1990s. We offer two illustrative applications, each quantifying causal changes in NTL as the result of a shock. We estimate changes in NTL emitted by 1) the city of Banda Aceh, Indonesia as a result of the 2004 tsunami and subsequent recovery efforts, and 2) the capital city of Hugo Chavez’s home state of Barinas, Venezuela after his ascension to Venezuela’s Presidency in 1999.
Geographic Disadvantage and the Composition of Trade (with Manuel Jimenez)
Landlocked countries and small islands face transport-related constraints on their ability to trade goods internationally. Such constraints can limit their ability to participate in international supply chains and to develop through international trade. This paper
seeks to shed light on the effects of transport-related constraints by relating the composition of countries’ exports and imports to characteristics associated with the method and cost of transoceanic transportation and to an indicator of products’ proximity
to final demand. We adopt an estimating strategy used to study levels of trade, but use it to study the composition of exports and imports. Estimated coefficients on the interactions between product- and country-characteristics reveal the way in which revealed comparative advantage varies with country characteristics. Relative to comparison countries, small islands’ and landlocked countries’ trade flows are biased towards products with particular transport characteristics. Among several new stylized facts, we find that both country-types’ imports and landlocked countries’ exports have below average weight-to-value ratios. Landlocked countries’ trade is biased against air-shipped products, and towards products that move in container vessels. Small islands’
imports are biased against products that move by air transport and by container vessels, a finding suggesting that small islands have comparative advantage in processing imported raw materials, most notably fuels. There is important heterogeneity of these and other such effects within each type of geography. The results offer broad lessons that can be informative for governments making transport infrastructure investments or other trade policies that favor one sector over another.
Landlocked countries and small islands face transport-related constraints on their ability to trade goods internationally. Such constraints can limit their ability to participate in international supply chains and to develop through international trade. This paper
seeks to shed light on the effects of transport-related constraints by relating the composition of countries’ exports and imports to characteristics associated with the method and cost of transoceanic transportation and to an indicator of products’ proximity
to final demand. We adopt an estimating strategy used to study levels of trade, but use it to study the composition of exports and imports. Estimated coefficients on the interactions between product- and country-characteristics reveal the way in which revealed comparative advantage varies with country characteristics. Relative to comparison countries, small islands’ and landlocked countries’ trade flows are biased towards products with particular transport characteristics. Among several new stylized facts, we find that both country-types’ imports and landlocked countries’ exports have below average weight-to-value ratios. Landlocked countries’ trade is biased against air-shipped products, and towards products that move in container vessels. Small islands’
imports are biased against products that move by air transport and by container vessels, a finding suggesting that small islands have comparative advantage in processing imported raw materials, most notably fuels. There is important heterogeneity of these and other such effects within each type of geography. The results offer broad lessons that can be informative for governments making transport infrastructure investments or other trade policies that favor one sector over another.
Geography, Transport and the Composition of Trade (with Manuel Jimenez and Bilgehan Karabay)
We develop a general equilibrium model of international trade and find support for it empirically. In the model a global transport sector allocates a fixed stock of shipping capital across bilateral routes. Sectors vary in the intensity in which they use transport capital. With intuition matching that of von Thunen’s model of regional agricultural specialization around an exogenous city, our model predicts that geographically central/large countries specialize in transport-intensive sectors, while remote regions specialize in goods that are less transport-intensive. The model improves on von Thunen’s because it is a complete general equilibrium. Our empirical results demonstrate that larger and more central regions specialize in products that rely on air shipment and shipment on containerized vessels. Remote regions export products that rely on less costly forms of shipping capital (e.g. bulk vessels). The results suggest an important potential constraint on remote regions’ ability to grow by producing increasingly complex goods.
We develop a general equilibrium model of international trade and find support for it empirically. In the model a global transport sector allocates a fixed stock of shipping capital across bilateral routes. Sectors vary in the intensity in which they use transport capital. With intuition matching that of von Thunen’s model of regional agricultural specialization around an exogenous city, our model predicts that geographically central/large countries specialize in transport-intensive sectors, while remote regions specialize in goods that are less transport-intensive. The model improves on von Thunen’s because it is a complete general equilibrium. Our empirical results demonstrate that larger and more central regions specialize in products that rely on air shipment and shipment on containerized vessels. Remote regions export products that rely on less costly forms of shipping capital (e.g. bulk vessels). The results suggest an important potential constraint on remote regions’ ability to grow by producing increasingly complex goods.
Economic consequences of cabotage restrictions: The effect of the Jones Act on Puerto Rico (with Manuel Jimenez)
We study the consequences of a US cabotage law for Puerto Rico (PR). The fleet of US vessels calling in PR lacks capacity for carrying non-containerized freight. PR’s imports of sea-shipped final products are biased against US mainland sources. This bias is strongest for heavy products and products not typically shipped in containers. Among upstream products, a strong bias against imports of sea-shipped products applies to all sources. Estimated tariff-equivalent costs among final products imply static annual welfare losses of 1.1 percent of household consumption ($203 per person). We present suggestive evidence that the policy generates additional dynamic losses.
We study the consequences of a US cabotage law for Puerto Rico (PR). The fleet of US vessels calling in PR lacks capacity for carrying non-containerized freight. PR’s imports of sea-shipped final products are biased against US mainland sources. This bias is strongest for heavy products and products not typically shipped in containers. Among upstream products, a strong bias against imports of sea-shipped products applies to all sources. Estimated tariff-equivalent costs among final products imply static annual welfare losses of 1.1 percent of household consumption ($203 per person). We present suggestive evidence that the policy generates additional dynamic losses.
Variable Scaling and Hypothesis Testing in the Gravity Model (with Anton Yang)
Statistical inference around hypothesis tests in Poisson Pseudo Maximum Likelihood (PPML) models is sensitive to data scaling choices. We show this analytically and demonstrate it using a simple application of the gravity model of trade. The scale of the data on the independent variable affects the scale of both the Likelihood statistic and the Likelihood Ratio test statistic. Lagrange Multiplier tests are similarly sensitive to data scaling choices. When considering Wald tests, we find some nuance. Data scaling affects the Wald statistic when it depends upon the asymptotic variance-covariance matrix, but not when the variance-covariance matrix depends upon residuals from robust estimation. Testing of joint hypotheses in PPML gravity models should therefore rely on Wald tests constructed from robust standard errors.
Statistical inference around hypothesis tests in Poisson Pseudo Maximum Likelihood (PPML) models is sensitive to data scaling choices. We show this analytically and demonstrate it using a simple application of the gravity model of trade. The scale of the data on the independent variable affects the scale of both the Likelihood statistic and the Likelihood Ratio test statistic. Lagrange Multiplier tests are similarly sensitive to data scaling choices. When considering Wald tests, we find some nuance. Data scaling affects the Wald statistic when it depends upon the asymptotic variance-covariance matrix, but not when the variance-covariance matrix depends upon residuals from robust estimation. Testing of joint hypotheses in PPML gravity models should therefore rely on Wald tests constructed from robust standard errors.
Renewable resource rents, taxation and the effects of wind power on rural economies (with Nhu "Claire" Nguyen)
The rapid growth of utility-scale wind energy generation is a potentially important boon to rural economies in the United States. Yet econometric estimates suggest that the local economic benefits of wind energy generation have been modest, perhaps because the sector is capital-intensive and financed almost exclusively by external capital. In this paper we argue that a) both the presence of a critical - but unpaid - factor of production (the wind) and generous federal subsidies are quantitatively important sources of economic rent, and b) a large portion of these rents accrue to providers of capital who reside outside the local economy. We build a partial equilibrium model that illustrates the mechanisms that generate economic rent, and integrate it into a small open economy general equilibrium model of a county’s economy. We calibrate the partial and general equilibrium models to data from two rural counties in Indiana, quantify the economic rents, and consider the consequences of a resource rent tax. Resource rent taxes generate significantly larger economic benefits for communities that host wind power, and offer an opportunity to spread the sector’s economic benefits more broadly within them. Broadly distributed revenues from resource rent taxes might facilitate greater acceptance of utility scale wind power in communities where the sector would otherwise be unwelcome. State public utility commissions provide an analytical infrastructure that could support local taxation of the kind that we consider.
The rapid growth of utility-scale wind energy generation is a potentially important boon to rural economies in the United States. Yet econometric estimates suggest that the local economic benefits of wind energy generation have been modest, perhaps because the sector is capital-intensive and financed almost exclusively by external capital. In this paper we argue that a) both the presence of a critical - but unpaid - factor of production (the wind) and generous federal subsidies are quantitatively important sources of economic rent, and b) a large portion of these rents accrue to providers of capital who reside outside the local economy. We build a partial equilibrium model that illustrates the mechanisms that generate economic rent, and integrate it into a small open economy general equilibrium model of a county’s economy. We calibrate the partial and general equilibrium models to data from two rural counties in Indiana, quantify the economic rents, and consider the consequences of a resource rent tax. Resource rent taxes generate significantly larger economic benefits for communities that host wind power, and offer an opportunity to spread the sector’s economic benefits more broadly within them. Broadly distributed revenues from resource rent taxes might facilitate greater acceptance of utility scale wind power in communities where the sector would otherwise be unwelcome. State public utility commissions provide an analytical infrastructure that could support local taxation of the kind that we consider.
Expediting Trade: Impact Evaluation of an In-House Clearance Program (with Ana M. Fernandes and Claudia Berg)
Despite the importance of trade facilitation as an area of trade and development policy, there have been very few impact evaluations of specific trade facilitation reforms. This paper offers an evaluation of in-house clearance, a reform that allows qualified firms in Serbia to clear customs from within their own warehouse rather than at the customs office. The pooled synthetic control method applied here offers a novel solution to many of the empirical challenges that frustrate efforts to evaluate trade facilitation reforms. The method is used to estimate causal impacts on trade outcomes for 21 firms that adopted in-house clearance for import shipments. The program compressed the distribution of clearance times for adopting firms, but the estimated effects on median clearance times, inspection rates, and import value were not statistically significant. Tests for heterogeneous program impact do not indicate that the program affected adopting firms differently. Overall, the results suggest that the most evident benefit of the program for participating firms is reduced uncertainty about clearance times.
Despite the importance of trade facilitation as an area of trade and development policy, there have been very few impact evaluations of specific trade facilitation reforms. This paper offers an evaluation of in-house clearance, a reform that allows qualified firms in Serbia to clear customs from within their own warehouse rather than at the customs office. The pooled synthetic control method applied here offers a novel solution to many of the empirical challenges that frustrate efforts to evaluate trade facilitation reforms. The method is used to estimate causal impacts on trade outcomes for 21 firms that adopted in-house clearance for import shipments. The program compressed the distribution of clearance times for adopting firms, but the estimated effects on median clearance times, inspection rates, and import value were not statistically significant. Tests for heterogeneous program impact do not indicate that the program affected adopting firms differently. Overall, the results suggest that the most evident benefit of the program for participating firms is reduced uncertainty about clearance times.
In Progress
Long-run Impacts of Incidental Trade Agreements (with Hao Xiong)
Long-run Impacts of Incidental Trade Agreements (with Hao Xiong)
