RESEARCH

Publications


Presentations

  • EMG Workshop, Sydney, Australia, 2019
  • Society for Economics Measurement Conference, Frankfurt, Germany, 2019
  • Productivity Research Network Fourth Conference, Sydney, Australia, 2019
  • EMG Workshop, Sydney, Australia, 2018
  • Forum of Chinese Energy and Environmental Economists, Xiamen, China, 2017
  • Systems of Engineering Society of China Annual Conference, Beijing, China, 2016

Working Papers

  • “Frontier firms, inefficiency and productivity dynamics” (Job Market Paper)

Productivity dynamics occur when firms enter and exit a market. Contributions from firms to industry productivity can be decomposed into effects from entrants, exits and incumbents. As opposed to productivity dynamics, productivity can also be decomposed into explanatory factors regarding efficiency and technical progress. These two patterns of decomposition provide different perspectives about the driving components of productivity. I propose a framework that merges them and produces a cross dimension. Industry productivity can not only be allocated as firm contributions, but also its explanatory factors can be illustrated analogously. It is developed by specifying firms that are on production frontiers, measuring the deviation from frontiers, and integrating explanatory factors with firm dynamics. A difference-in-differences approach is proposed that validates the firm dynamics from the counterfactual perspective. As an empirical exercise, the framework is applied to Australian firm-level data and reveals the dominant contribution of incumbent firms to industry productivity and industry efficiency.

  • “Hedonic imputation with tree-based decision approaches”

Linear hedonic regression is commonly utilised to estimate missing prices of unmatched products, but the linear assumption of prices in product characteristics is dubious. Actual consumer purchase patterns show that product characteristics are not perfectly substitutable so that the prediction capacity of linear models is challenged. I consider alternative estimations of hedonic prices by introducing tree-based machine learning models that are highly recommended for prediction accuracy. A tree decision structure is compatible with consumer preferences when product characteristics are complements. Model performance metrics from (electronic-point-of-sale) scanner data confirm the prediction accuracy of tree-based models. I find that random forests are the best fitted model with largest R2-type measures among a series of tree-based models. Price indexes with random forests display correct predictions that are robust in single, double and full imputation. The variable importance estimated for product characteristics is consistent with real coefficients of hedonic functions in price simulation. It is advisable that tree-based decision approaches, especially random forests, can be effectively employed for unmatched products in hedonic imputation.

  • “Industry-level value added and productivity decompositions”

Decomposing productivity into explanatory factors has been extensively used for identifying the responsible components of economic growth. Drawing on a non-parametric model, this paper decomposes value added and productivity into explanatory factors for Australian market sector. 12 selected industries and 16 market sector industries are analysed, featuring an industry-level decomposition. The results indicate that technical progress has supported the increasing productivity, though it is slightly offset by input mix factor and value added inefficiency. The effect of share weighting is validated when computing contributions to overall performance from industries. Financial and insurance services contribute to overall productivity due to large market shares and technical progress while the mining industry tends to have the largest negative impact on efficiency and leads to overall productivity deterioration. The priority of efficiency is empirically confirmed for industry production in Australia.

Create your website with WordPress.com
Get started
<span>%d</span> bloggers like this: