The Determinants of Cross-Border Investment: A Value-Chain Analysis (joint with Marta Noguer)
"la Caixa" Working Paper No 5. First Draft: December 2006.

This paper contributes to the literature on the determinants of FDI. We use a new data set covering greenfield and expansion projects to examine which factors influence the decision to invest abroad. Our empirical framework is an augmented gravity model that incorporates elements of factor proportions theory. At the aggregate level, we find that distance discourages FDI, size and sharing a language encourages it, and that FDI targets relatively capital scarce countries. When we classify investment projects according to their stage in the chain of production, we observe that the effect of all the factors vary substantially across stages. Furthermore, even though the results confirm FDI targetting capital scarce countries, we find evidence of a minimum requirement on the host country´s capital endowment in all the stages of production except extraction. Finally, education is very important for FDI in R&D and regional HQ and easiness of doing business is especially important for the later.

Outsourcing and your Collar's Color , Submitted
"la Caixa" Working Paper No 3. First Draft: March 2006. This Draft: November 2006.

In order to assess the effects of increased outsourcing on the relative demand for skilled and unskilled labor, it is crucial to understand whether outsourcing is a complement or a substitute for each kind of labor. Using the traditionally employed log-log framework, Amiti and Wei (2006) find that outsourcing of goods and labor are complements. Using the same methodology but differentiating between skilled and unskilled labor, one would conclude that outsourcing acts as a complement to unskilled labor but a substitute for skilled labor. This paper proposes an improved methodology which uses estimated prices for outsourcing instead of other proxies (such as its intensity) and a complete factor cost-share system of equations to find the completely opposite result, that is, outsourcing is a substitute for unskilled labor and a complement for skilled labor. This result is consistent with the findings of the literature on outsourcing and the wage gap.

What Explains the Widening Wage Gap? Outsourcing vs. Technology , Submitted
"la Caixa" Working Paper No 1. First Draft: September 2005. This Draft: November 2006.

The relative rise of wages for high-skilled workers over the last three decades has been the subject of intense academic and popular scrutiny. This paper develops a new methodology for decomposing wage changes into four sources: outsourcing, biased technological change, total technological change, and international price movements. We find that between 1980 and 1999 outsourcing accounts for 28 percent of the observed wage change, and biased technological change for another 18 percent. Jointly these two forces explain 58 percent of the wage change. The remaining 42 percent can be explained by international price changes. In sum, we find that outsourcing and biased technological change can account for a large share of the observed divergence in the skilled wage premium.

Trade Patterns, Trade Balances, and Idiosyncratic Shocks (joint with Xavier Gabaix, Josep M Vilarrubia, David E. Weinstein) 
Working Paper, April 2005.

International Macroeconomics has long sought an explanation for current account fluctuations that matches the data. The approaches have typically focused on better models and new macroeconomic variables. We demonstrate the limitations of this approach by showing that idiosyncratic firm level shocks are important cause of macroeconomic volatility even for large countries. When explaining these fluctuations, standard macroeconomic models generally assume that firms are small and that their microeconomic shocks cancel out. We show that the high degree of concentration of bilateral trade flows means that idiosyncratic shocks can have a significant impact on aggregate economic fluctuations. We theoretically develop a decomposition of the variance of trade flows into its macroeconomic and its microeconomic components. Taking the model to data on bilateral trade flows from 1970 to 1997, we find that the most comprehensive macroeconomic model can only account for at most half of the observed variance in trade account volumes of each country. Thus, this paper highlights the importance of considering disaggregated data when modelling the current account.

Assessing the Short Run Consequences of Bereavement on Medical Expenditures in the Elderly Population (joint with Mitali Das), Revision Health Services Research.

A long line of research in social psychology, as well as vast anecdotal evidence, has documented the negative health consequences of the death of a spouse. Surprisingly little, however, is known about the health or medical expenditures of the surviving spouse subsequent to bereavement. The objective of this paper is to empirically assess whether there are any discernible effects of bereavement on out-of-pocket prescription expenditures in the elderly population in the immediate aftermath of spousal bereavement, and whether such effects impose burdensome costs on any strata of the elderly. This question is a useful point to consider in the ongoing debate on increasing prescription coverage under Medicare. We find that spousal bereavement is indeed associated with a increase in prescription expenditures. This result is sharpest for nonwhites, smaller for widows (white and nonwhite) and absent for white widowers. In the initial months after bereavement, monthly prescription expenditures for the bereaved may increase by 22-30% of pre-bereavement levels, and correspond with approximately one additional prescription filled every 2 months. These effects are small for the median-income elderly, but for the bereaved elderly individual on Medicare who fills more than 25% of the average number of prescriptions and is in the lowest-fourth of the income distribution, total prescription expenditures due to both bereavement and prior conditions could be as much as 13% of income.

Nonclassical Measurement Error in Nonrandomly Selected Data (joint with Mitali Das and Changxiu Li) , Revision Journal of the American Statistical Association.

The analysis of models with measurement error in binary valued covariate has been of considerable interest, as it raises problems quite distinct form measurement error in continuously-distributed covariates. However, in many data where such measurement error arises, there is another important and perhaps overlooked source of bias: the non-random selection of the observed data. Accounting for measurement error in a binary covariate in such instances is in general insufficient to obtain consistent parameter estimates. Compounding this problem is that the true binary covariate appears in both the selection and outcome equations. We consider this problem when a noisy report of the true variable is available, and derive a point identification for all parameters of interest. In contrast to previous work on nonclassical measurement error, our methods require only a single noisy report. We illustrate our methods by analyzing the effects of illicit maternal drug usage on infant birthweight. Here we show the large parameter biases in ignoring either measurement error in drug-usage report, selection of birthweight, or both.