Presented at: NOITS 2025; University of Stavanger; 47th Annual Meeting of the Norwegian Association of Economists; Kristiania University of Applied Sciences; and University of Oslo (scheduled)
Abstract: This paper estimates the elasticity of substitution between clean and dirty energy, a central parameter for evaluating climate policy, yet remains scarce and widely dispersed in the empirical literature. I show that much of this dispersion is mechanical: the two dominant identification strategies recover the same object only under the joint restrictions of (i) Constant Elasticity of Substitution (CES) and (ii) Hicks-neutral technology (HN). Price-based designs identify a single constant slope from relative prices under CES, while marginal-product designs back out elasticities from estimated production functions under NH; once either restriction fails, the two approaches generally diverge. I derive the exact source of this divergence and develop a bridging estimator that relaxes both CES and HN, jointly recovering firm–time elasticities and factor-biased productivities. The two conventional methods are nested as testable special cases. Applying the estimator to Norwegian administrative data on food-and-beverage manufacturing (2010–2022), I find near-Leontief energy aggregation: the clean–dirty elasticity clusters close to zero, implying limited short-run within-firm fuel switching in response to relative-price changes. I cannot reject CES, and CES-based estimates closely match my unrestricted results. In contrast, HN is rejected; imposing it leads marginal-product approaches to overstate substitutability and can even suggest gross substitutes.
A Green Trade-Off? Agricultural Technical Change and Deforestation (with Torfinn Harding). Link.
Presented at: NOITS 2025; BI Norwegian Business School; Swedish University of Agricultural Sciences; University of Stavanger; 47th Annual Meeting of the Norwegian Association of Economists; Kristiania University of Applied Sciences; OsloMet; Umeå University
Abstract: This paper studies the adoption of a labor-saving technology, and its effects on the adoption of a complementary land-saving technology and on land use. We exploit the legalization of labor-saving genetically engineered (GE) soybeans in Brazil as a natural experiment. GE-soy adoption is higher in municipalities with greater natural soy potential, and initial labor scarcity amplifies this relationship. The resulting reduction in labor demand in the soy sector creates a positive labor supply shock for other sectors. This facilitates the adoption of the land-saving technology of double-cropping (DC) maize, which adds a second harvest season. We find that the two technologies are complementary in the longer run, enabling the parallel expansion of soy and maize production with limited cropland expansion and deforestation.
Presented at: EARIE 2025; 46th annual meeting of the Norwegian Association of Economists; NAERE Workshop; University of Stavanger
Abstract: General equilibrium models often rely on nested CES production functions to simulate energy and environmental policies. Yet these functions typically impose rigid substitution structures for tractability rather than empirical validity. This paper introduces a new framework for estimating flexible, multi-level CES functions using observable revenue and input data, without solving difficult equilibrium conditions. We also propose two tools for selecting the most empirically supported nesting structure: one based on significance tests of transformation terms, and another using testable restrictions within translog models. Using simulations and input-output data, we show that widely used CES structures misrepresent actual substitution patterns, especially in settings involving renewable and fossil energy.
Presented at: University of Stavanger
Abstract: There is a gap in the production function estimation literature, where factor-biased productivities and factor misallocations are typically examined separately. Factor-biased productivity studies often exclude misallocation by making specific market assumptions, while research on misallocation tends to ignore factor-biased productivity. We argue that this separation likely stems from an underidentification challenge that arises when both factor-biased technical change and misallocation are present simultaneously. Building on the recent frameworks, we propose a robust production function estimation method to address this issue and predict the biases inherent in standard approaches that account for only a subset of productivities and misallocations. To confirm the validity of our method and the predicted biases, we conduct extensive Monte Carlo simulations.