Publications
Bias and Consistency in Three-way Gravity Models (w/ Martin Weidner)
Journal of International Economics, 132. 2021. (appendix)
We study the incidental parameter problem for the three-way PPML estimator recently recommended for identifying the effects of trade policies. We show that, while PPML has unique consistency properties in this context, it suffers from an asymptotic bias whenever the time dimension is fixed, and cluster-robust standard error estimates are generally biased as well. We also present several bias-correction methods that lead to more reliable inferences.
Currency Unions and Trade: A PPML Re-Assessment with High-Dimensional Fixed Effects
(w/ Mario Larch, Joschka Wanner, and Yoto Yotov)
Oxford Bulletin of Economics and Statistics, 81(3), 487-510 (lead article). 2019. (appendix)
In principle, rigorous trade policy analysis requires large data sets, ideally using as many countries and years as possible. However, large samples tend to limit choice of estimator, leaving an important methodological gap. We address this gap by unveiling a high-dimensional fixed effects PPML estimator suitable for trade policy analysis with comprehensive trade data. Our PPML estimates of the currency union effect and of the effect of the Euro specifically offer very different conclusions than an otherwise rigorously-specified linear model.
Do Cross-border Patents Promote Trade? (w/ Claire Brunel)
Canadian Journal of Economics, 55(1), 379-418. 2022.
We perform a systematic panel data analysis of foreign patents and trade. We find that, yes, patents evidently do promote trade. But perhaps not in the way you might think: among other things, we show that patents have had significantly larger effects in less-differentiated industries where monopoly rights over a particular variety should be less valuable.
The Effects of Free Trade Agreements on Product-level Trade (w/ Scott French)
European Economic Review, 162. 2024. (appendix)
We test whether "least traded products" experience faster trade growth after the signing of trade agreements and whether this is important for the aggregate trade creation effects of trade agreements. The answer is yes and yes.
Firm Heterogeneity and Trade-induced Layoffs: An Empirical Investigation (w/ Pinar Uysal and Yoto V. Yotov)
European Economic Review, 75, 80-97. 2015. (working paper version)
We use a novel data set with verified observations of trade-induced layoffs by U.S. firms to study how firm productivity and exposure to tariff reductions jointly determine the number of job losses due to trade. Guided by the predictions of the Melitz (2003) heterogeneous firms theory, we find the number of trade-induced layoffs decreases with firm productivity for exporting firms, but actually increases with productivity for non-exporters.
On the trade diversion effects of Free Trade Agreements (w/ Mian Dai and Yoto V. Yotov)
Economics Letters, 122(2), 321-325. 2014.
Our novel gravity estimation approach uses pair-wise variation in “outside FTA's”—agreements either trading partner has with a third country—to confirm that FTAs divert trade away from non-member countries. Interestingly, we find that diversion from internal trade (domestic sales) due to FTAs has been significantly stronger than diversion from external trade.
On the Widely Differing Effects of Free Trade Agreements: Lessons from Twenty Years of Trade Integration
(w/ Scott Baier and Yoto Yotov)
Journal of International Economics, 116, 206-226. 2019. (working paper version)
We develop a novel two-stage methodology that allows us to study and predict heterogeneity in the effects of FTAs. We offer several intuitive ways of explaining the rich variation we observe in our estimates of FTA effects. Even within the same agreement, FTA effects are weaker for more distant pairs, but stronger for those with otherwise high levels of ex ante trade frictions. Guided by theory, we also relate asymmetric effects within pairs to each country’s ability to influence the other’s terms of trade. We demonstrate the usefulness of our methods by simulating the effects of TTIP.
ppmlhdfe: Fast Poisson Estimation with High-Dimensional Fixed Effects (with Sergio Correia and Paulo Guimarães)
Stata Journal. 20(1), 95-115. 2020. (working paper version)
We present a new algorithm for estimating Poisson regressions w/ high-dimensional fixed effects based on iteratively reweighted least squares (IRLS). Our algorithm takes advantage of recent methods for fast estimation of large linear models as well as some new techniques we introduce for accelerating IRLS w/ high-dimensional covariates. We also include a comprehensive check for the non-existence of estimates and incorporate remedies from a companion paper.
Prudence versus Predation and the Gains from Trade (with Michelle Garfinkel and Constantinos Syropoulos)
Journal of Economic Theory, 201. 2022. (appendix)
We use a two-country dynamic model to study how the perceived likelihood of a future international conflict shapes resource allocation decisions and preferences over trade in the present. Our theory demonstrates that the larger of the two countries could find trade to be unappealing due to the ensuing consequences for its future national security.
Journal of International Economics, 132. 2021. (appendix)
We study the incidental parameter problem for the three-way PPML estimator recently recommended for identifying the effects of trade policies. We show that, while PPML has unique consistency properties in this context, it suffers from an asymptotic bias whenever the time dimension is fixed, and cluster-robust standard error estimates are generally biased as well. We also present several bias-correction methods that lead to more reliable inferences.
- See here for a Stata package that implements the methods described in this paper.
Currency Unions and Trade: A PPML Re-Assessment with High-Dimensional Fixed Effects
(w/ Mario Larch, Joschka Wanner, and Yoto Yotov)
Oxford Bulletin of Economics and Statistics, 81(3), 487-510 (lead article). 2019. (appendix)
In principle, rigorous trade policy analysis requires large data sets, ideally using as many countries and years as possible. However, large samples tend to limit choice of estimator, leaving an important methodological gap. We address this gap by unveiling a high-dimensional fixed effects PPML estimator suitable for trade policy analysis with comprehensive trade data. Our PPML estimates of the currency union effect and of the effect of the Euro specifically offer very different conclusions than an otherwise rigorously-specified linear model.
- Replaces older working paper: "The Currency Union Effect: A PPML Reassessment with High-Dimensional Fixed Effects"
Do Cross-border Patents Promote Trade? (w/ Claire Brunel)
Canadian Journal of Economics, 55(1), 379-418. 2022.
We perform a systematic panel data analysis of foreign patents and trade. We find that, yes, patents evidently do promote trade. But perhaps not in the way you might think: among other things, we show that patents have had significantly larger effects in less-differentiated industries where monopoly rights over a particular variety should be less valuable.
The Effects of Free Trade Agreements on Product-level Trade (w/ Scott French)
European Economic Review, 162. 2024. (appendix)
We test whether "least traded products" experience faster trade growth after the signing of trade agreements and whether this is important for the aggregate trade creation effects of trade agreements. The answer is yes and yes.
Firm Heterogeneity and Trade-induced Layoffs: An Empirical Investigation (w/ Pinar Uysal and Yoto V. Yotov)
European Economic Review, 75, 80-97. 2015. (working paper version)
We use a novel data set with verified observations of trade-induced layoffs by U.S. firms to study how firm productivity and exposure to tariff reductions jointly determine the number of job losses due to trade. Guided by the predictions of the Melitz (2003) heterogeneous firms theory, we find the number of trade-induced layoffs decreases with firm productivity for exporting firms, but actually increases with productivity for non-exporters.
On the trade diversion effects of Free Trade Agreements (w/ Mian Dai and Yoto V. Yotov)
Economics Letters, 122(2), 321-325. 2014.
Our novel gravity estimation approach uses pair-wise variation in “outside FTA's”—agreements either trading partner has with a third country—to confirm that FTAs divert trade away from non-member countries. Interestingly, we find that diversion from internal trade (domestic sales) due to FTAs has been significantly stronger than diversion from external trade.
On the Widely Differing Effects of Free Trade Agreements: Lessons from Twenty Years of Trade Integration
(w/ Scott Baier and Yoto Yotov)
Journal of International Economics, 116, 206-226. 2019. (working paper version)
We develop a novel two-stage methodology that allows us to study and predict heterogeneity in the effects of FTAs. We offer several intuitive ways of explaining the rich variation we observe in our estimates of FTA effects. Even within the same agreement, FTA effects are weaker for more distant pairs, but stronger for those with otherwise high levels of ex ante trade frictions. Guided by theory, we also relate asymmetric effects within pairs to each country’s ability to influence the other’s terms of trade. We demonstrate the usefulness of our methods by simulating the effects of TTIP.
- Featured in CESifo Newsletter and VoxEU.
- Recognized by JIE as their most cited article since 2018.
ppmlhdfe: Fast Poisson Estimation with High-Dimensional Fixed Effects (with Sergio Correia and Paulo Guimarães)
Stata Journal. 20(1), 95-115. 2020. (working paper version)
We present a new algorithm for estimating Poisson regressions w/ high-dimensional fixed effects based on iteratively reweighted least squares (IRLS). Our algorithm takes advantage of recent methods for fast estimation of large linear models as well as some new techniques we introduce for accelerating IRLS w/ high-dimensional covariates. We also include a comprehensive check for the non-existence of estimates and incorporate remedies from a companion paper.
- github repository with examples, code, and benchmarks
- Verifying the Existence of Maximum Likelihood Estimates for Generalized Linear Models, a companion paper that describes the problem of non-existence of estimates in nonlinear GLM models.
Prudence versus Predation and the Gains from Trade (with Michelle Garfinkel and Constantinos Syropoulos)
Journal of Economic Theory, 201. 2022. (appendix)
We use a two-country dynamic model to study how the perceived likelihood of a future international conflict shapes resource allocation decisions and preferences over trade in the present. Our theory demonstrates that the larger of the two countries could find trade to be unappealing due to the ensuing consequences for its future national security.
Working Papers
Bootstrap for Gravity Models (slides) (preliminary)
Feeding China's Rise: The Growth Effects of Trading with China, 1993-2011
How much has China's trade expansion contributed to world economic growth over the last two decades? I investigate this question using a dynamic trade and growth model that combines endogenous capital accumulation with a rich sectoral dimension, where each sector differs in its usage as an input to consumption, investment, and production. The “dynamic sectoral linkages” highlighted by this setting deliver a wider set of possibilities for the gains from trade and, ultimately, larger real income gains from China's trade expansion, especially in the long run.
Machine Learning in International Trade Research - Evaluating the Impact of Trade Agreements (with Holger Breinlich, Valentina Corradi, Nadia Rocha, Michele Ruta, and João Santos Silva) (slides)
We combine the three-way gravity model often used to study the effects of trade agreements with a lasso method for selecting the most important predictors from a set of high-dimensional covariates. Using comprehensive data on the provision content of trade agreements, we find that provisions related to anti-dumping, competition policy, TBTs, and trade facilitation are most relevant for promoting trade.
The Problem of Peace: A Story of Corruption, Destruction, and Rebellion (with Constantinos Syropoulos)
We explore how the central presence of state institutions in civil conflicts generates unique explanations for the emergence of destructive war. We do so in a model where a kleptocratic government and an equally self-interested “rebel” rival compete for insecure resources by raising armies from a common labor population. We show the government may actively favor conflict in this setting, even though conflict results in wasteful destruction, in order to enhance the value it derives from its use of fiscal policy.
Verifying the Existence of Maximum Likelihood Estimates for Generalized Linear Models
(w/ Sergio Correia and Paulo Guimarães)
There has been considerable ambiguity over how to verify whether estimates from nonlinear models “exist” and what can be done if they do not. We characterize the problem in detail across a wide range of generalized linear models and introduce a novel method for dealing with it in the presence of high-dimensional fixed effects, as are often recommended for gravity models and in other common panel settings.
Feeding China's Rise: The Growth Effects of Trading with China, 1993-2011
How much has China's trade expansion contributed to world economic growth over the last two decades? I investigate this question using a dynamic trade and growth model that combines endogenous capital accumulation with a rich sectoral dimension, where each sector differs in its usage as an input to consumption, investment, and production. The “dynamic sectoral linkages” highlighted by this setting deliver a wider set of possibilities for the gains from trade and, ultimately, larger real income gains from China's trade expansion, especially in the long run.
Machine Learning in International Trade Research - Evaluating the Impact of Trade Agreements (with Holger Breinlich, Valentina Corradi, Nadia Rocha, Michele Ruta, and João Santos Silva) (slides)
We combine the three-way gravity model often used to study the effects of trade agreements with a lasso method for selecting the most important predictors from a set of high-dimensional covariates. Using comprehensive data on the provision content of trade agreements, we find that provisions related to anti-dumping, competition policy, TBTs, and trade facilitation are most relevant for promoting trade.
- Featured in The Economics of Deep Trade Agreements and VoxEU.
The Problem of Peace: A Story of Corruption, Destruction, and Rebellion (with Constantinos Syropoulos)
We explore how the central presence of state institutions in civil conflicts generates unique explanations for the emergence of destructive war. We do so in a model where a kleptocratic government and an equally self-interested “rebel” rival compete for insecure resources by raising armies from a common labor population. We show the government may actively favor conflict in this setting, even though conflict results in wasteful destruction, in order to enhance the value it derives from its use of fiscal policy.
Verifying the Existence of Maximum Likelihood Estimates for Generalized Linear Models
(w/ Sergio Correia and Paulo Guimarães)
There has been considerable ambiguity over how to verify whether estimates from nonlinear models “exist” and what can be done if they do not. We characterize the problem in detail across a wide range of generalized linear models and introduce a novel method for dealing with it in the presence of high-dimensional fixed effects, as are often recommended for gravity models and in other common panel settings.
- "Revise and resubmit" at Econometric Reviews
- github repository with examples, code, and benchmarks
- ppmlhdfe: Fast Poisson Estimation with High-Dimensional Fixed Effects, a companion paper that describes our new ppmlhdfe Stata command.
Code
ge_gravity: Stata program for solving a simple general equilibrium trade model. Ideal for bootstrapping GE results based on prior gravity estimates. (program) (help file) (examples)
penppml: R package for penalized (lasso and ridge) PPML regressions with high-dimensional fixed effects.
ppmlhdfe: Poisson Pseudo-maximum Likelihood estimation with high-dimensional fixed effects. For use in Stata.
ppml_fe_bias: Bias corrections for Poisson Pseudo-maximum Likelihood gravity models with two- and three-way fixed effects. For use in Stata. (program) (help file) (examples)
- Suggested citation: “On the Widely Differing Effects of Free Trade Agreements” (see above)
- Also now available as an R package (w/ thanks to Vadim Kudlay)
penppml: R package for penalized (lasso and ridge) PPML regressions with high-dimensional fixed effects.
- Suggested citation: "Machine Learning in International Trade Research" (see above)
ppmlhdfe: Poisson Pseudo-maximum Likelihood estimation with high-dimensional fixed effects. For use in Stata.
- Suggested citations (2): “Verifying existence” and the “ppmlhdfe” companion paper (see above)
ppml_fe_bias: Bias corrections for Poisson Pseudo-maximum Likelihood gravity models with two- and three-way fixed effects. For use in Stata. (program) (help file) (examples)
- Suggested citation: "Bias and Consistency in Three-way Gravity Models" (see above)
- Featured in the WTO's "An Advanced Guide to Trade Policy Analysis".
- Suggested citation: “Currency Unions and Trade” (see above)
More about me...
Citizenship: UK/US (dual national)
Education: University of Pennsylvania (B.A., Mathematics and English, 2005); Drexel University (Ph.D., Economics, 2015)
Hometown: Wilmington, DE
Education: University of Pennsylvania (B.A., Mathematics and English, 2005); Drexel University (Ph.D., Economics, 2015)
Hometown: Wilmington, DE