Finite sample properties of Moran's I test for spatial autocorrelation in tobit models

Pedro V. Amaral, Luc Anselin

Research output: Contribution to journalArticle

15 Citations (Scopus)

Abstract

In this note, we investigate the finite-sample properties of Moran's I test statistic for spatial autocorrelation in tobit models suggested by Kelejian and Prucha. We fill a void in the theoretical literature by investigating the finite sample properties of this test statistic in a series of Monte Carlo simulations, using data sets ranging from 49 to 15,625 observations. We find that the test is unbiased, has considerable power and approximates the asymptotic normal distribution even for medium-sized sample sizes, empirically confirming the theoretical results of Kelejian and Prucha. However, some caution is needed, since the statistic turns out to be sensitive to misspecification in the form of heteroscedasticity. In such instances the test over-rejects the null hypothesis, mistaking heteroscedasticity for spatial autocorrelation.

Original languageEnglish (US)
Pages (from-to)773-781
Number of pages9
JournalPapers in Regional Science
Volume93
Issue number4
DOIs
StatePublished - Nov 1 2014

Fingerprint

autocorrelation
statistics
void
simulation
test

Keywords

  • C21
  • C25
  • Moran's
  • Spatial econometrics
  • Spatial tobit

ASJC Scopus subject areas

  • Environmental Science (miscellaneous)
  • Geography, Planning and Development

Cite this

Finite sample properties of Moran's I test for spatial autocorrelation in tobit models. / Amaral, Pedro V.; Anselin, Luc.

In: Papers in Regional Science, Vol. 93, No. 4, 01.11.2014, p. 773-781.

Research output: Contribution to journalArticle

Amaral, Pedro V. ; Anselin, Luc. / Finite sample properties of Moran's I test for spatial autocorrelation in tobit models. In: Papers in Regional Science. 2014 ; Vol. 93, No. 4. pp. 773-781.
@article{c63003f57e7a46fd95d59ed5a5112a34,
title = "Finite sample properties of Moran's I test for spatial autocorrelation in tobit models",
abstract = "In this note, we investigate the finite-sample properties of Moran's I test statistic for spatial autocorrelation in tobit models suggested by Kelejian and Prucha. We fill a void in the theoretical literature by investigating the finite sample properties of this test statistic in a series of Monte Carlo simulations, using data sets ranging from 49 to 15,625 observations. We find that the test is unbiased, has considerable power and approximates the asymptotic normal distribution even for medium-sized sample sizes, empirically confirming the theoretical results of Kelejian and Prucha. However, some caution is needed, since the statistic turns out to be sensitive to misspecification in the form of heteroscedasticity. In such instances the test over-rejects the null hypothesis, mistaking heteroscedasticity for spatial autocorrelation.",
keywords = "C21, C25, Moran's, Spatial econometrics, Spatial tobit",
author = "Amaral, {Pedro V.} and Luc Anselin",
year = "2014",
month = "11",
day = "1",
doi = "10.1111/pirs.12034",
language = "English (US)",
volume = "93",
pages = "773--781",
journal = "Papers in Regional Science",
issn = "1056-8190",
publisher = "Wiley-Blackwell",
number = "4",

}

TY - JOUR

T1 - Finite sample properties of Moran's I test for spatial autocorrelation in tobit models

AU - Amaral, Pedro V.

AU - Anselin, Luc

PY - 2014/11/1

Y1 - 2014/11/1

N2 - In this note, we investigate the finite-sample properties of Moran's I test statistic for spatial autocorrelation in tobit models suggested by Kelejian and Prucha. We fill a void in the theoretical literature by investigating the finite sample properties of this test statistic in a series of Monte Carlo simulations, using data sets ranging from 49 to 15,625 observations. We find that the test is unbiased, has considerable power and approximates the asymptotic normal distribution even for medium-sized sample sizes, empirically confirming the theoretical results of Kelejian and Prucha. However, some caution is needed, since the statistic turns out to be sensitive to misspecification in the form of heteroscedasticity. In such instances the test over-rejects the null hypothesis, mistaking heteroscedasticity for spatial autocorrelation.

AB - In this note, we investigate the finite-sample properties of Moran's I test statistic for spatial autocorrelation in tobit models suggested by Kelejian and Prucha. We fill a void in the theoretical literature by investigating the finite sample properties of this test statistic in a series of Monte Carlo simulations, using data sets ranging from 49 to 15,625 observations. We find that the test is unbiased, has considerable power and approximates the asymptotic normal distribution even for medium-sized sample sizes, empirically confirming the theoretical results of Kelejian and Prucha. However, some caution is needed, since the statistic turns out to be sensitive to misspecification in the form of heteroscedasticity. In such instances the test over-rejects the null hypothesis, mistaking heteroscedasticity for spatial autocorrelation.

KW - C21

KW - C25

KW - Moran's

KW - Spatial econometrics

KW - Spatial tobit

UR - http://www.scopus.com/inward/record.url?scp=84908894905&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84908894905&partnerID=8YFLogxK

U2 - 10.1111/pirs.12034

DO - 10.1111/pirs.12034

M3 - Article

AN - SCOPUS:84908894905

VL - 93

SP - 773

EP - 781

JO - Papers in Regional Science

JF - Papers in Regional Science

SN - 1056-8190

IS - 4

ER -