Regression and Inference Under Smoothness Restrictions

Similar documents
Nonparametric Estimates of the Clean and Dirty Energy Substitutability

Estimation of growth convergence using a stochastic production frontier approach

Using Non-parametric Methods in Econometric Production Analysis: An Application to Polish Family Farms

1. The OLS Estimator. 1.1 Population model and notation

Public and Private Capital Productivity Puzzle: A Nonparametric Approach

CONSTRAINED NONPARAMETRIC KERNEL REGRESSION: ESTIMATION AND INFERENCE

Review of Classical Least Squares. James L. Powell Department of Economics University of California, Berkeley

Nonparametric and Semiparametric Regressions Subject to Monotonicity Constraints: Estimation and Forecasting

Environmental Regulation and U.S. State-Level Production

Cost Efficiency, Asymmetry and Dependence in US electricity industry.

7) Important properties of functions: homogeneity, homotheticity, convexity and quasi-convexity

ECONOMETRIC THEORY. MODULE VI Lecture 19 Regression Analysis Under Linear Restrictions

DEPARTMENT OF ECONOMICS

LECTURE 1- PRODUCTION, TECHNOLOGY AND COST FUNCTIONS (BASIC DEFINITIONS)-EFFICIENCY AND PRODUCTIVITY MEASUREMENT

Hendrik Wolff University of Washington. July 2013

Review of Econometrics

Partial derivatives BUSINESS MATHEMATICS

Imposing and Testing for Shape Restrictions in. Flexible Parametric Models 1

Specification Testing of Production in a Stochastic Frontier Model

Exploring County Truck Freight. By : Henry Myers

Statistics and econometrics

Chapter 4 Differentiation

A Note on the Scale Efficiency Test of Simar and Wilson

CONSTRAINED NONPARAMETRIC KERNEL REGRESSION: ESTIMATION AND INFERENCE

A FLEXIBLE TIME-VARYING SPECIFICATION OF THE TECHNICAL INEFFICIENCY EFFECTS MODEL

Lecture 3: Multiple Regression

Intermediate Econometrics

Choice is Suffering: A Focused Information Criterion for Model Selection

ECON 186 Class Notes: Optimization Part 2

Public Sector Management I

An Information Theoretic Approach to Flexible Stochastic Frontier Models

Econometrics I Lecture 3: The Simple Linear Regression Model

The profit function system with output- and input- specific technical efficiency

Practical Econometrics. for. Finance and Economics. (Econometrics 2)

SOLUTIONS Problem Set 2: Static Entry Games

The outline for Unit 3

LECTURE NOTES ON MICROECONOMICS

SINGLE-STEP ESTIMATION OF A PARTIALLY LINEAR MODEL

Estimation of Theoretically Consistent Stochastic Frontier Functions in R

Dynamics of Dairy Farm Productivity Growth. Johannes Sauer

Returns-to-Scaleto Marginal product describe the

Introduction Large Sample Testing Composite Hypotheses. Hypothesis Testing. Daniel Schmierer Econ 312. March 30, 2007

A Bootstrap Test for Conditional Symmetry

Applied Econometrics (QEM)

Bagging Nonparametric and Semiparametric Forecasts with Constraints

Heteroskedasticity-Robust Inference in Finite Samples

Bootstrap Testing in Econometrics

Density estimation Nonparametric conditional mean estimation Semiparametric conditional mean estimation. Nonparametrics. Gabriel Montes-Rojas

Functional Form. Econometrics. ADEi.

Comparing Bayesian Networks and Structure Learning Algorithms

Geometric intuition of least squares Consider the vector x = (1, 2) A point in a two-dimensional space

Flexible Estimation of Treatment Effect Parameters

Econ 583 Final Exam Fall 2008

Semiparametric Cost Allocation Estimation

The Logit Model: Estimation, Testing and Interpretation

Independent and conditionally independent counterfactual distributions

Multiple Sample Categorical Data

Tutorial 3: Optimisation

School of Business. Blank Page

Lecture: Simultaneous Equation Model (Wooldridge s Book Chapter 16)

Workshop I The R n Vector Space. Linear Combinations

Non-linear panel data modeling

Nonparametric Analysis of Cost Minimization and Efficiency. Beth Lemberg 1 Metha Wongcharupan 2

FENG CHIA UNIVERSITY ECONOMETRICS I: HOMEWORK 4. Prof. Mei-Yuan Chen Spring 2008

Derivative Properties of Directional Technology Distance Functions

1 Mixed effect models and longitudinal data analysis

Quick Review on Linear Multiple Regression

Final Exam. Economics 835: Econometrics. Fall 2010

Lecture 13: Subsampling vs Bootstrap. Dimitris N. Politis, Joseph P. Romano, Michael Wolf

Shape constrained kernel-weighted least squares: Estimating production functions for Chilean manufacturing industries

ECON2285: Mathematical Economics

Mathematical Economics: Lecture 9

Birkbeck Working Papers in Economics & Finance

This paper is not to be removed from the Examination Halls

A Nonparametric Kernel Representation of the Agricultural Production Function: Implications for Economic Measures of Technology

Discussion of the paper Inference for Semiparametric Models: Some Questions and an Answer by Bickel and Kwon

METHODOLOGY AND APPLICATIONS OF. Andrea Furková

Weighted Least Squares

Input-biased technical progress and the aggregate elasticity of substitution: Evidence from 14 EU Member States

Stochastic Nonparametric Envelopment of Data (StoNED) in the Case of Panel Data: Timo Kuosmanen

Nonlinear Inequality Constrained Ridge Regression Estimator

Nonparametric Estimation of Homothetic and Homothetically Separable Functions. Boston College Yale University, October, 2004

Wild Bootstrap Inference for Wildly Dierent Cluster Sizes

Partial Differentiation

Hakone Seminar Recent Developments in Statistics

Multiple comparisons of slopes of regression lines. Jolanta Wojnar, Wojciech Zieliński

Shape-Constrained Kernel-Weighted Least Squares: Estimating Production Functions for Chilean Manufacturing Industries

Tangent Plane. Nobuyuki TOSE. October 02, Nobuyuki TOSE. Tangent Plane

Nonstationary Panels

Advanced Microeconomics

Interpreting Regression Results

Comparative Statics. Autumn 2018

Answers to Problem Set #4

DOCUMENTS DE TRAVAIL CEMOI / CEMOI WORKING PAPERS

Nonparametric Estimation in a One-Way Error Component Model: A Monte Carlo Analysis

IMPOSING REGULARITY CONDITIONS ON A SYSTEM OF COST AND FACTOR SHARE EQUATIONS. William E. Griffiths, Christopher J. O'Donnell and Agustina Tan Cruz

Gibbs Sampling in Endogenous Variables Models

Math Review ECON 300: Spring 2014 Benjamin A. Jones MATH/CALCULUS REVIEW

A COMPARISON OF HETEROSCEDASTICITY ROBUST STANDARD ERRORS AND NONPARAMETRIC GENERALIZED LEAST SQUARES

Transcription:

Regression and Inference Under Smoothness Restrictions Christopher F. Parmeter 1 Kai Sun 2 Daniel J. Henderson 3 Subal C. Kumbhakar 4 1 Department of Agricultural and Applied Economics Virginia Tech 2,3,4 Department of Economics Binghamton University The 44th. Annual Conference of the CEA

Introduction Research Question Motivation How to impose generalized economic constraints on parametric model for the estimation of a single/multiple-output technology? Most of the applied econometrics literature on the estimation of technology ignores the imposition of economic constraints. Negative marginal products or technical regress Various methods developed to impose constraints

Research Question Motivation Motivation Production Function Example Cobb-Douglas ln Y = α + β K ln K + β L ln L + u ln Y / ln K = β K 0 MPK 0 Translog ln Y =α + β K ln K + β L ln L + 0.5β KK (ln K) 2 + 0.5β LL (ln L) 2 + β KL ln K ln L + u ln Y / ln K = β K + β KK ln K + β KL ln L 0 MPK 0 ln Y / ln K is now observation-specific.

Motivation Introduction Research Question Motivation Recent work on constrained regression Hall and Huang (2001) O Donnell, Rambaldi and Doran (2001) Racine, Parmeter and Du (2009) Henderson and Parmeter (2009) What is new? Complement O Donnell et al. (2001) with classical approach Extend Racine et al. (2009) to any linear estimator

A Simple Simulation Graphical Illustration Data Generating Process: y = 10 + 3x + x 2 3x 3 + x 4 + u where x iidu[0, 2.5] u iidn(0, 0.01) Estimate ŷ = ˆβ 0 + ˆβ 1 x + ˆβ 2 x 2 + ˆβ 3 x 3 + ˆβ 4 x 4 Constraint: ŷ/ x 0.

Without Constraint Introduction A Simple Simulation Graphical Illustration

With Constraint Introduction A Simple Simulation Graphical Illustration

Constrained Regression: Estimation Constrained Regression: Estimation Constrained Regression: Inference Idea: Transform the response so that certain equality or inequality constraints are satisfied. Y = Xβ + u; Ŷ = Xb, where X = [1 x x 2 x 3 x 4 ] The unconstrained gradient: Ŷ(X)/ x = ( X/ x) b = [0 1 2x 3x 2 4x 3 ] (X X) 1 X Y n = x (X X) 1 X Y = A(x)Y = A i (x)y i The constrained gradient: i=1 Ŷ(X p)/ x = n i=1 A i(x)p i Y i = n i=1 A i(x)y i 0.

Constrained Regression: Estimation Constrained Regression: Inference Constrained Regression: Estimation Weight selection criterion p that minimizes D(p) = (p u p) (p u p) subject to Ŷ(X p)/ x 0 p: a weighting vector for the response [p 1, p 2,..., p n ] p u : a uniform weighting vector If p = p u, constraints will be non-binding.

Constrained Regression: Estimation Constrained Regression: Inference Constrained Regression: Inference Null hypothesis: Constraints are valid. A bootstrap approach: 1. Estimate the model under constraints, get D(ˆp), Ŷ(X ˆp), û; 2. Bootstrap û; obtain u ; 3. Create a new response, Y = Ŷ(X ˆp) + u ; 4. Estimate the model using the new sample under constraints, obtain D(ˆp ); 5. Repeat step 2 to 4 (B) times, calculate ˆP B = 1 B B b=1 I(D(ˆp ) D(ˆp)).

Econometric Model Data Results Future Work Multiple-output Technology Cannot use a production function Y: output vector (Q 1); X : input vector (K 1) Start with transformation function f (Y, X, t) = 1 or ln f (Y, X, t) = 0 ln f (Y, X, t) f (ln Y, ln X, t)

Econometric Model Data Results Future Work Multiple-output Technology Impose the restriction of homogeneous of degree 1 in inputs, and use X 1 (the first input) as the normalizing input: ln f (Y, X, t) ln X 1 f (ln Y, ln X, t) where ln X = ln(x /X 1 ) is a (K 1)-vector Estimate: ln X 1 = f (ln Y, ln X, t, id ) + u where id is a categorical variable for state effects f ( ): either a Translog (parametric) or an unknown smooth function (kernel-based nonparametric as in Li and Racine (2006)).

Econometric Model Data Results Future Work Multiple-output Technology Input distance function Shephard (1953) Färe and Primont (1995) Kumbhakar and Lovell (2000) Economic constraints: ln X 1 / ln X k 0 ( k = 2,..., K) ln X 1 / ln Y q 0 ( q = 1,..., Q) ln X 1 / t 0

Econometric Model Data Results Future Work Data: United States Department of Agriculture Table: Summary Statistics of the Variables Symbol Variable Name Mean Sd. Min. Max. X 1 Capital 638932 579797 7351 3330621 X 2 Land 681420 734737 4015 4659258 X 3 Labor 1616318 1409715 18189 8450988 X 4 Intermediate inputs 1915233 1753558 12917 9451845 Y 1 Livestock 1813097 1697585 9101 8497604 Y 2 Crop 2355130 2695164 25216 19386468 Y 3 Agricultural services 237087 309676 981 26603678 1. All of the variables are measured as real index numbers. 2. The sample consists of 1200 observations (for 48 states and 25 yearly data (1980 2004) for each state).

Econometric Model Data Results Future Work Estimation Results (Translog)

Econometric Model Data Results Future Work Estimation Results (Nonparametric)

Econometric Model Data Results Future Work Hypothesis Testing Seven joint monotonicity constraints Technical progress Constant returns to scale

Econometric Model Data Results Future Work Future Work Impose non-linear constraints on linear estimator Impose linear constraints on a seemingly-unrelated regression (SUR) system Extend the methodology to non-linear estimator