An improved Afriat-Diewert-Parkan nonparametric production function estimator

Ole Bent Olesen, John Ruggiero

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

14 Downloads (Pure)

Resumé

Recent developments in the production frontier literature include nonparametric estimators with shape constraints. A few of these estimators rely on the Afriat inequalities to provide piecewise linear approximations to the production function/frontier. We show in this paper that these Afriat–Diewert–Parkan (ADP) estimators have deficiencies in the presence of moderate statistical noise including overfitting and a relatively high estimator variance. We propose new estimators with lower variance and a relatively low bias. We consider such alternative estimators based on (weighted) averages of random hinge functions with parameter restrictions. Small sample properties of the estimators are presented that show our new estimators outperform the existing ADP estimators when moderate to large amounts of noise are present.

OriginalsprogEngelsk
TidsskriftEuropean Journal of Operational Research
Vol/bind264
Udgave nummer3
Sider (fra-til)1172-1188
ISSN0377-2217
DOI
StatusUdgivet - 1. feb. 2018

Fingeraftryk

Production Function
Estimator
Hinges
Shape Constraint
Piecewise Linear Approximation
Variance Estimator
Overfitting
Random Function
Nonparametric Estimator
Weighted Average
Production function
Small Sample
Restriction
Alternatives

Citer dette

@article{bfdaf11c92e54993bb43a61fd155f6e8,
title = "An improved Afriat-Diewert-Parkan nonparametric production function estimator",
abstract = "Recent developments in the production frontier literature include nonparametric estimators with shape constraints. A few of these estimators rely on the Afriat inequalities to provide piecewise linear approximations to the production function/frontier. We show in this paper that these Afriat–Diewert–Parkan (ADP) estimators have deficiencies in the presence of moderate statistical noise including overfitting and a relatively high estimator variance. We propose new estimators with lower variance and a relatively low bias. We consider such alternative estimators based on (weighted) averages of random hinge functions with parameter restrictions. Small sample properties of the estimators are presented that show our new estimators outperform the existing ADP estimators when moderate to large amounts of noise are present.",
keywords = "Concave nonparametric frontier estimators, Data envelopment analysis, Hinge functions, Model averaging, Stochastic DEA",
author = "Olesen, {Ole Bent} and John Ruggiero",
year = "2018",
month = "2",
day = "1",
doi = "10.1016/j.ejor.2017.07.057",
language = "English",
volume = "264",
pages = "1172--1188",
journal = "European Journal of Operational Research",
issn = "0377-2217",
publisher = "Elsevier",
number = "3",

}

An improved Afriat-Diewert-Parkan nonparametric production function estimator. / Olesen, Ole Bent; Ruggiero, John.

I: European Journal of Operational Research, Bind 264, Nr. 3, 01.02.2018, s. 1172-1188.

Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

TY - JOUR

T1 - An improved Afriat-Diewert-Parkan nonparametric production function estimator

AU - Olesen, Ole Bent

AU - Ruggiero, John

PY - 2018/2/1

Y1 - 2018/2/1

N2 - Recent developments in the production frontier literature include nonparametric estimators with shape constraints. A few of these estimators rely on the Afriat inequalities to provide piecewise linear approximations to the production function/frontier. We show in this paper that these Afriat–Diewert–Parkan (ADP) estimators have deficiencies in the presence of moderate statistical noise including overfitting and a relatively high estimator variance. We propose new estimators with lower variance and a relatively low bias. We consider such alternative estimators based on (weighted) averages of random hinge functions with parameter restrictions. Small sample properties of the estimators are presented that show our new estimators outperform the existing ADP estimators when moderate to large amounts of noise are present.

AB - Recent developments in the production frontier literature include nonparametric estimators with shape constraints. A few of these estimators rely on the Afriat inequalities to provide piecewise linear approximations to the production function/frontier. We show in this paper that these Afriat–Diewert–Parkan (ADP) estimators have deficiencies in the presence of moderate statistical noise including overfitting and a relatively high estimator variance. We propose new estimators with lower variance and a relatively low bias. We consider such alternative estimators based on (weighted) averages of random hinge functions with parameter restrictions. Small sample properties of the estimators are presented that show our new estimators outperform the existing ADP estimators when moderate to large amounts of noise are present.

KW - Concave nonparametric frontier estimators

KW - Data envelopment analysis

KW - Hinge functions

KW - Model averaging

KW - Stochastic DEA

U2 - 10.1016/j.ejor.2017.07.057

DO - 10.1016/j.ejor.2017.07.057

M3 - Journal article

VL - 264

SP - 1172

EP - 1188

JO - European Journal of Operational Research

JF - European Journal of Operational Research

SN - 0377-2217

IS - 3

ER -