Skip to main content
Skip to McMaster Navigation Skip to Site Navigation Skip to main content
McMaster logo
COVID-19 information and updates

Find the most recent updates here, as well as FAQs and information for students, faculty and staff.

Kealey John

John Kealey

Department of Economics


Fields of Specialization

Primary: Applied Econometrics, Applied Nonparametric Statistics
Secondary: Firm Productivity, International Trade


  • 2014-2016 - SSHRC Doctoral Scholarship
  • 2013-2014 - Ontario Graduate Scholarship


  • 2015 - “Distributional trends in firm productivity before and after trade liberalization”,
    Canadian Economics Association Annual Conference
  • 2016 - “Nonparametric estimation of stochastic production frontier models for panel
    data”, Canadian Economics Association Annual Conference, University of Toronto
    Doctoral Workshop in Applied Econometrics


Ph.D. Economics, McMaster University, expected 2016

M.A. International Affairs, Carleton University, 2011

M.A. Economics, Dalhousie University, 2009

B.A. Economics, University of Victoria, 2007


  • 2013 (Winter) - Teaching Assistant, Intermediate Microeconomics II, McMaster University
  • 2012 (Fall) - Teaching Assistant, Intermediate Microeconomics II, McMaster University
  • 2012 (Winter) - Teaching Assistant, Intermediate Microeconomics I, McMaster University
  • 2011 (Fall) - Teaching Assistant, Intermediate Microeconomics I, McMaster University
  • 2009 (Winter) - Teaching Assistant, Principles of Microeconomics, Dalhousie University
  • 2009 (Fall) - Teaching Assistant,Principles of Microeconomics, Dalhousie University


Publications and Papers

Robust nonparametric frontier estimation for continuous or count data

Abstract: This paper proposes a robust nonparametric estimation procedure for
deterministic frontier models with either a continuous or a discrete count-valued output
variable. It exhibits minimal sensitivity to outliers without resorting to aggressive data
trimming that tends to have an overreaching effect on non-extreme-valued
observations. The estimator's favourable performance is primarily attributable to its use
of a novel trimming parameter selection routine that combines k-means and hierarchical
clustering techniques. Evidence from a Monte Carlo experiment suggests that both the
nonsmooth and the smooth versions of the proposed estimator give rise to a better fit of
the frontier function than existing robust data envelopment methods. The paper
concludes with an empirical example that uses historical patent count data from the
U.S. manufacturing sector to estimate the efficiency of firm-level R & D spending.

Nonparametric estimation of stochastic production frontier models for panel data

Abstract: This paper proposes a nonparametric estimation procedure for stochastic
production frontier models for panel data, making it possible to decompose firm-level
inefficiency into a persistent and a time-varying component. Existing approaches tend to
rely on a simple linear specification of the frontier function, but this can give rise to
imprecise estimates of i) the frontier itself, ii) factor elasticities, and iii) firm-level
inefficiency. In contrast, the proposed method dispenses with parametric assumptions
vis-a-vis the functional form of the frontier and the distribution of inefficiency, thereby
avoiding some of the potentially adverse consequences of model misspecification.
Using panel data from the Colombian manufacturing sector, it is shown that the kernel-based
approach is better able to account for heterogeneity in production technology that
exists across firms and over time. Most importantly, the nonparametric framework yields
substantially different estimates of firm-level inefficiency than its parametric counterpart;
in fact, in many cases, the two classes of estimates are characterized by a first-order
stochastic dominance relationship.

Semiparametric estimation of a Cobb-Douglas production function with varying elasticity

Abstract: This paper proposes a semiparametric method of estimating a Cobb-Douglas
model of firm-level production in which the elasticity coefficients can be functions of both
continuous and discrete predictors. It is shown that the varying-coefficient method is
better able to reflect the hypothesized relationship between factor elasticities and their
corresponding input expenditure shares. Using plant-level data from the Colombian
manufacturing sector, an empirical example is provided in which the elasticity of output
with respect to capital and labour can vary by industry and across different time periods.
In this setting, the contribution of unskilled labour to final output diminishes over time,
which brings about a rather sharp decline in estimated returns to scale.