Research
Working Papers
Exporting State-Promoted Technologies and the Direction of Global Innovation (Job Market Paper)
Abstract: Policymakers have recognized the importance of information technologies as a strategic area for innovation (e.g. AI, 5G, semiconductors). These technologies are often governed by technical standards developed at consensus-based standard-setting organizations (SSOs) which are led by private sector firms with incentives to promote their own technologies. As part of political agenda, Chinese firms participating in international SSOs are asked to collectively promote certain domestic technologies for the industry standards. How does a China-led standard shape the subsequent innovation globally? Specifically: Are the promoted technologies the “best” solutions for the industry? If adopted as the standard, do firms build on these technologies in their new 5G inventions? I answer these questions focusing on 5G wireless communication technology and its standardization at an international consortium of SSOs called the 3rd Generation Project Partnership (3GPP). I build a novel dataset based on thousands of 3GPP official documents, China’s policy documents describing 5G technologies identified as national goals, and patents that I identify as 5G technologies. I find that a China-led standard has implications for both static and dynamic inefficiencies. China promotes “catch-up” technologies in which China is less advanced than other countries but filed many patents over the past years. The Chinese firms at the 3GPP act collectively to make them industry standards in the presence of better alternatives (static). With an IV strategy, I find that a China-led standard causes both foreign (leader) and Chinese (laggard) innovators to innovate more building on these suboptimal solutions for 5G. That is, the “previous” leader is pulled down to compete with the “successful” laggard generating duplicated innovation (dynamic). Neither China nor any other country responds to an adoption of optimal solutions as the standard (e.g. US, European technologies). A central coordination in a consensus-based institution generates a non-market environment allowing the laggard to be successful.
Global Input Sourcing and the Role of Supplier Risks
Abstract: I build a trade model featuring cross-country differences in cost-risk profiles to evaluate the role of supplier risks in firms’ input sourcing decisions. The model provides an intuitive characterization of the optimal sourcing strategy which depends on the risk-adjusted cost-competitiveness and the relative diversification benefits of the countries in a firm's choice set. In the model, multi-sourcing can arise as risk averse firms choose to diversify away risks. I use the model to empirically assess the role of risk diversification in three exercises: 1. Using the model calibrated to represent US firms’ sourcing decisions, I show that firms’ motives to diversify risks can explain 30% of the observed variation in import shares for intermediate goods; 2. In an exercise replicating the rise in US tariffs on Chinese intermediate goods during 2018-2019, I find that risk diversification leads to stronger responses by US firms in terms of their sourcing strategies through more pronounced extensive margin adjustments and additional adjustments along the intensive margin; 3. In a counterfactual transition from autarky to the observed trade equilibrium, US firms diversify partly through higher-cost countries resulting in higher marginal costs which are not fully offset by lower markups charged for risk-taking.
Works in Progress
Geography of Trade Diversification and International Transmission of Shocks (with Jay Hyun and Gaelan MacKenzie)
Pre-Doctoral Publication/Work
Abstract: Using panel data covering all firms performing R&D in Canada, we estimate the external return to R&D by size of firm, defining the spillover pool using a measure of technological proximity based on firms' reported expenditure in 147 research fields. Our study confirms that the spillover benefits from R&D are substantial, so government support for R&D is justified. However, we find that spillovers rise with the size of R&D performers, so Canada's policy of subsidizing R&D performed by small firms at a higher rate is not warranted. We also find a much lower private rate of return on R&D for small firms than for large firms.
Op-ed in the Globe and MailCited in the Financial Post
Abstract: A key economic issue in Canada is the declining business research and development and slow-down in the total factor productivity (TFP) growth in manufacturing since the early 2000s. To deepen our understanding of this phenomenon, we focus on the increasing Chinese import share in the total domestic absorption in Canadian manufacturing since the early 2000s, which appears to be driven by positive supply shocks within Chinese manufacturing. Based on firm-level data covering all incorporated firms in Canadian manufacturing, we find that rising Chinese import competition led to declines in R&D expenditure and TFP growth within firms but reallocated employment towards more productive firms and induced less productive firms to exit. The negative within-effects were pronounced for firms that were initially smaller, less profitable, and less productive. At the aggregate level, the positive reallocation effects on TFP more than offset the negative within-effect. We estimate that, had there been no increase in Chinese import competition between 2005 and 2010, TFP in Canadian manufacturing would have declined by 1.26 per cent per year instead of the actual 1.09 per cent per year over this period.
Abstract: In this paper, we describe our Stata program thresvar for the estimation of threshold vector autoregressive models (TVAR) and conducting the associated linearity test. For the latter, we carry out a Monte-Carlo simulation study on the finite-sample distribution of the test statistics for the inference. The study is motivated by the fact that the threshold parameters are nuisance parameters which are not identified under the null of linearity. Thus, the asymptotic distribution of the test statistics is non-standard and potentially affected by the unidentified parameters. Based on our findings, we suggest that one relies on a simulation method for inference on a case-by-case basis. Lastly, we present a simple example in which we estimate a two-regime TVAR using our thresvar command based on the U.S. quarterly data for real GDP, inflation and BAA spread.
Op-ed: There's a plus side of the China Shock for Canadians (with Andrew Sharpe)In Financial Post