Search results
Ernest Liu. Assistant Professor of Economics. My research interests are in networks, growth, trade, finance, and macro-development. I am affiliated with Princeton's Bendheim Center for Finance, the International Economics Section, and the Julis-Rabinowitz Center for Public Policy and Finance. CV (PDF)
- Contact
Accessibility Help; © 2024 The Trustees of Princeton...
- Cas
By proceeding to access and use University computing and...
- Research
Reject and resubmit, American Economic Review. “Neoclassical...
- Innovation Networks and R&D Allocation
Ernest Liu† Princeton and NBER Song Ma‡ Yale and NBER...
- Contact
Reject and resubmit, American Economic Review. “Neoclassical Growth in an Interdependent World”, with Benny Kleinman, Stephen Redding, and Moto Yogo. "Innovation Networks and R&D Allocation", with Song Ma.
Ernest Liu is a versatile performer who has appeared in films, TV shows and web series. He is best known for his roles in From Dusk Till Dawn, Chimères and Jesus.
- 32 sec
Articles 1–20. Princeton University - Cited by 1,166 - Growth - Finance - Macroeconomics - Development Economics.
- 1 Introduction
- Preferences and Production Technology
- 2.2 Optimal Allocation of R&D Resources
- 2.3 R&D Allocation and Economic Growth
- s.t. b ≥
- 2.5 General Functional Forms and Endogenous Innovation Network
- 2.6 Knowledge Spillovers from Abroad
- 2.7.3 Potential Ineficiency in A Decentralized Market
- 3 Data
- 4 Innovation Network and Knowledge Spillovers
- Innovation Centrality Across Sectors
- 5.6 Innovation Hubs
- A.3 Proof of Lemma 2: Economic Growth Rate Along a Balanced Growth Path
- A.5 Proof of Proposition 3: Welfare Impact of R&D Reallocation
- A. Proof of Proposition 6: Optimal R&D in the Presence of Foreign Spillovers
- + ̄gu) du
- B.4 An Illustrative Decentralized Equilibrium
- B.5 Constrained Optimal R&D Allocations
- Ω ◦ Xf .
- ˆγK
- C Details on Data Construction
- C.1 U.S. Patent Data
- C.3 Connecting Patent Data with Sectoral Data
- D Cross-checking Google Patents with PATSTAT
- D.4 Robustness of Results Using Google Patents and PATSTAT
- E Supplementary Results
How to foster innovation has long been a central question for economists and policy makers. The discussion has concentrated on the amount of resources invested in research and develop-ment (R&D) and the cost of under- or over-investment. But how should these R&D resources be allocated across economic sectors or technological fields? This question i...
There is a representative consumer with log flow utility and exponential discounting at rate ρ:
In this section we characterize the optimal allocation of R&D resources in the economy. Consider a benevolent social planner who chooses the entire time path of worker and scientist allocations across sectors to maximize consumer utility. We can write the planner’s problem as
In this section we show how R&D allocation afects the economic growth rate along a balanced growth path (BGP). We demonstrate that the network’s eigenvector centrality—what we call “in-novation centrality”—is a suficient statistic for evaluating the growth rate along a BGP and coin-cides with the growth-maximizing R&D allocation. We show the social...
This corollary highlights that innovation centrality a coincides with the growth-maximizing R&D allocation along a BGP. Intuitively, ai captures the extent to which sector i’s R&D activities contribute to economic growth, taking into account the network efects. Sectors with higher innovation centrality represent more fundamental technologies in the...
The baseline model features an exogenous innovation network Ω, as the elasticities of each sec- ∂ tor’s innovation productivity to another sector’s knowledge stock (ωij lnχit ≡ ) are exogenous ∂ lnqjt structural parameters. The knowledge spillover dynamics (9) in the baseline model thus form a log-linear dynamical system. Such log-linearity—along w...
We will later use our model to assess R&D allocations in real-world economies. As we show, some countries, like the U.S. and Japan, rely more on domestic knowledge spillovers and less on foreign knowledge spillovers, while other economies benefit more from foreign spillovers particularly from the technologically advanced ones. We now extend our mod...
Why may a decentralized market not allocate R&D resources eficiently? In an innovation net-work, knowledge is a public good, as knowledge creation benefits subsequent R&D in other sec-tors and all future periods. To the extent that innovators do not fully internalize such future benefits,3 a decentralized market does not implement the eficient R&D ...
This section describes the data for our empirical analyses. We use patent citation data across sectors and countries to construct the global innovation network. We also use data on sectoral production, final use, and R&D. Here we briefly describe how we construct and harmonize these data. Section C of the Online Appendix provides more details.
In this section, we build several key data elements that will be used in our main quantitative analy-sis in Section 5. We first construct the innovation network Ω and discuss its empirical properties. We then empirically validate a key mechanism in our model, that knowledge spillovers occur through innovation networks both domestically in the U.S. ...
We provide some descriptive statistics of the innova-tion centrality a, which is the dominant left eigenvector of the innovation network Ω. Recall that
What explains cross-country diferences in R&D allocative eficiency? We do not have defini-tive answers, but we can present a conjecture with some empirical support: firms whose R&D activities span multiple sectors and technology classes allocate their resources in ways that may resemble the social planner’s. Because these firms’ R&D activities buil...
Consider a BGP in which R&D allocation shares follow the vector b and the growth rate of sectoral knowledge stock is time-invariant. The law of motion for stock vector is d ln qt/dt = λ · (ln η + ln ̄s + ln b + (Ω − I) ln qt). Taking derivative with respect to time,
The law of motion for knowledge stock ln q under R&D allocation b is
First, note that given output yt and the price of imports pf t , consumption, export, and import must solve C∗ ̄ yt, pf t ≡ max C cd , cf t t cd t,cf t
For given initial levels of knowledge stock and path of worker allocation, the diference in welfare under two R&D allocations eb and b is
In an innovation network, knowledge is a public good, as knowledge creation benefits subsequent R&D in other sectors and all future periods. To the extent that innovators do not fully internalize such future benefits,13a decentralized market does not implement the optimal R&D allocation. To demonstrate the potential ineficiency, in this section we ...
In some settings, for instance under political or feasibility constraints, a planner may only be able to reallocate resources across a subset K ⊂ {1, . . . , K} of sectors. We now generalize our results to such an environment. We show that our earlier results extend naturally: resources among sectors in K should be allocated proportionally to the u...
Optimal worker allocation should follow β, ˆ and optimal R&D allocation should follow ˆγ′ ρ ≡ ρ + λ Ω ˆ
i = γi ˆ . γj ˆ The consumption-equivalent welfare gains from adopting the optimal domestic R&D allocation (in-stead of allocation b) is LK (b) = exp ψλ ρ ˆγj ˆγK ′ ln ˆγK − ln b .
In this appendix, we provide details on data collection and harmonization and robustness of our approach.
U.S. patent data are obtained from the United States Patent and Trademark Ofice (USPTO).15 The data include information on patent inventors and patent assignee, allowing us to identify the geographic locations of the innovation (e.g., identifying cases in which a Chinese firm is granted a USPTO patent). We also observe the timing of the patents inc...
Patent data are classified into International Patent Classification (IPC) classes based on the tech-nological content of the invention. The IPC system provides a uniform and hierarchical system of language-independent symbols for the classification of patents and utility model according to the diferent areas of technology to which they pertain. The...
This appendix compares data from Google Patents (accessible to all researchers free of charge) and the widely used commercial database PATSTAT. These exercises will compare their data coverage, key variable definitions, and the robustness of empirical analyses in those two databases.
In this section, we present results from using PATSTAT patent data as the base for innovation measurement and innovation network construction. The overall takeaway is that the results using PATSTAT are virtually identical to results using Google Patents.
In this section, we provide additional empirical results.
Nov 10, 2023 · The Economics Department at Princeton University congratulates Ernest Liu, who has been awarded a 2023 Excellence Award from the Kiel Institute for the World Economy. Liu is an Assistant Professor of Economics in Princeton’s Economics Department who works in finance, networks, trade, growth, and macro-development.
Ernest Liu studies the effects of weak financial institutions on economic growth, allocation, and development. He uses production network theory, interest rate policies, and market power to analyze industrial policies and underinvestment in developing economies.
