Industrial policies are back in fashion: tariffs, subsidies, discriminatory regulations, and other interventions are sending shock waves through global supply chains. Indeed, a single supply chain can experience a multitude of shocks simultaneously: increased input costs from tariffs, decreased access to foreign markets due to foreign tariffs, and heightened competition from subsidized foreign counterparts. In this complex setting, I turn to quantitative general equilibrium modeling to answer two questions. First, what is a country’s optimal portfolio of tariffs and subsidies when, as in the real world, policies are targeted at specific industries in specific countries at specific points in the global supply chain? Second, what are the optimal policies and their impacts when, as happens in practice, unilateral trade actions invoke foreign retaliation? Answering these questions requires high-dimensional modeling of production networks, especially when global supply chains amplify policy impacts. I estimate the impact of the policy portfolio on output, trade, and welfare under three scenarios: (a) unilateral action where each country acts alone; (b) a multilateral Nash tariff and subsidy game that captures retaliation; and (c) a global central planner. Given the large number of policies in the portfolio, I use Shapley values to identify the most important ones. Four conclusions emerge. (1) Optimal policies targeting services, such as IT and finance, have the greatest impacts, indicating that it is not enough to study manufacturing alone. (2) Chinese optimal subsidy policy is more extensive and impactful than that of the United States. (3) In the Nash equilibrium, subsidies are far more welfare-damaging than optimal tariffs alone. (4) All tariffs and subsidies act like a nuclear option, setting off chain reactions that reverberate through supply networks and stress the economies of third parties.

Programming language: Python

Abstract:
Industrial policy has gained popularity in recent years and across all regions and income levels. Consequently, it is increasingly important to understand how governments choose the sectors they target. This analysis explores the role of domestic production networks in sector targeting, while controlling for other sector and global value chain characteristics. Combining datasets on industrial policy (Global Trade Alert) and input-output linkages (ICIO, OECD) provides novel insight into the network features of industrial policy. In particular, a sector’s ‘centrality’—i.e., its degree of connectedness – within the domestic production network is an important and significant predictor of sector intervention. The results indicate that industrial policy is used differently across regions, income groups, time periods, and types of policy tools. Notably, emerging economies tend to target more central sectors, while advanced economies target less central ones, on average. However, there has been a global shift toward more central sectors over time. Lastly, subsidies are deployed on more central sectors, while tariffs are used on less central ones.

Programming languages: Python, STATA

Abstract: Budgeting, an exemplar of good financial practices, informs financial decisions. This research employs the Theory of Reasoned Action to identify potential antecedents of individuals’ decisions to maintain a budget. Analysis of the 2014 Canadian Financial Capability Survey (CFCS) shows that the likelihood of budgeting depends on attitudes toward finance, reliance on professional advice, financial knowledge, and confidence. Women are relatively more likely to budget than men, but self-employed individuals are no more likely to have a budget than paid employees. The latter result is consistent with research findings of high rates of failure among young enterprises. The need for yet greater financial education is implied.

This paper builds a novel dataset combining public patent data and proprietary plant-level data on physical capital in the semiconductor industry. Using Natural Language Processing on patent abstracts I establish detailed firm technology profile over time. This is contrasted against location-specific firm investment in physical capital. The analysis presents novel empirical regularities about a key industry and evidence of the link between R&D investment and the location of physical capital.

Programming Languages: Python, SQL, STATA

In recent years, public bikesharing programs have become a common feature in many cities, offering residents and visitors a flexible transportation service for commuting, errand-running, and leisure. This analysis investigates the relationship between bikeshare and the traditional public transportation system. Using granular, trip-level data, I develop a non-parametric model to describe the use of bikeshare at a specific location. Compared to this baseline ridership, I find that severe subway delays are associated with significantly higher bikeshare use, while moderate delays are not.

Programming languages: R

Project presents a structural model of imperfect competition where the customer-facing firm simultaneously exerts markups over their customers and mark-downs over their suppliers. This market structure can be found in platform-based e-commerce services, as well as some traditional industries. The model demonstrates the need to jointly estimate markups, marginal costs, and markdowns. In the absence of sufficiently granular data, as is usually the case, this poses econometric challenges. The project presents simulated results.

Programming language: MATLAB

Processed and visualized multi-dimensional data on industrial policy interventions in preparation for departmental paper.

Performed independent research examining the role of production networks in determining industrial policy. Datasets on industrial policy and production networks are linked. Policies are placed in their production network context by computing centrality measures at domestic and global levels). Centrality is demonstrated as a key determinant of the decision to target a sector with subsidy/tariff. Project culminated in working paper published by IMF.

Programming languages: Python, STATA

Geolocated firms/institutions that hold AI patents. First, the location (coordinates) of each patent holder’s HQ is extracted using Wikipedia API. Then, using the coordinates, the address is retrieved using Google Maps API.
Programming language: Python

Project explores wealth accumulation along the life-cycle using longitudinal, administrative Norwegian data on asset holdings. Wrote code to perform econometric analysis remotely on administrative dataset housed at the Bank of Norway.

Programming language: STATA

Performed literature review.

Gathered data on venture capital rounds for clean tech and biotech firms.