Causal Relationship between Asset Prices and Output in the United States: Evidence from the State-Level Panel Granger Causality Test

Abstract

Emirmahmutoglu F., Balcilar M., Apergis N., Simo-Kengne B. D., Chang T. and Gupta R. Causal relationship between asset prices and output in the United States: evidence from the state-level panel Granger causality test, Regional Studies. This paper investigates the causal relationship between asset prices and output across US states using a bootstrap panel Granger causality approach which allows not only for heterogeneity and cross-sectional dependence to be accounted for but also interdependency between asset markets. Empirical results from a trivariate vector autoregression (VAR) comprising real house prices, real stock prices and real per capita personal income over 1975–2012 reveal the existence of a unidirectional causality running from both asset prices to output. This confirms the leading indicator property of asset prices for the real economy, while also substantiating the wealth and/or collateral transmission mechanism.

Emirmahmutoglu F., Balcilar M., Apergis N., Simo-Kengne B. D., Chang T. and Gupta R. 美国资产价格与产出之间的因果关係：州层级的面板格兰杰因果关係检定证据，区域研究。本文运用自助面板格兰杰因果关係方法，探讨美国各州的资产价格和产出之间的因果关係。该方法不仅能够考量异质性与跨部门依赖，亦可同时考量资产市场之间的相互依赖性。包含 1975 年至 2012 年间真实住宅价格、真实股票价格与真实人均所得的三元向量自迴归（VAR）模型的经验结果，显示出从资产价格到产出，单一方向的因果关係同时存在。此一结果，确认了资产价格之于实际经济所扮演的先导指标属性，并同时证实了财富与／或抵押传递机制。

Emirmahmutoglu F., Balcilar M., Apergis N., Simo-Kengne B. D., Chang T. et Gupta R. Le lien de causalité entre les prix des actifs et la production aux États-Unis: des résultats provenant du test de causalité au sens de Granger par panel à l’échelle de l’état, Regional Studies. Cet article examine le lien de causalité entre les prix des actifs et la production à travers les états aux É-U employant une méthode de causalité au sens de Granger par panel étendu au modèle bootstrap qui tient compte non seulement de l'hétérogénéité et de la dépendance transversale mais aussi de l'interdépendance entre les marchés des actifs. Les résultats empiriques provenant d'un modèle autorégressif multivarié (VAR) qui comprend les prix réels de l'immobilier, les prix réels des actifs et le revenu personnel réel par tête entre 1975 et 2012 laissent voir la présence d'une causalité unidirectionnelle allant des prix des actifs à la production. Cela confirme la caractéristique principale du prix des actifs comme indicateur de l’économie réelle, tout en justifiant également le mécanisme de transmission de la richesse et/ou du nantissememnt.

Emirmahmutoglu F., Balcilar M., Apergis N., Simo-Kengne B. D., Chang T. und Gupta R. Kausale Beziehung zwischen Vermögenspreisen und Produktion in den USA: Belege eines Granger-Kausalitätstests für ein Panel auf Bundesstaatsebene, Regional Studies. In diesem Beitrag wird die kausale Beziehung zwischen Vermögenspreisen und Produktion in den verschiedenen US-Bundesstaaten mithilfe eines grangerschen Bootstrap-Panel-Kausalitätsansatzes untersucht, der nicht nur eine Berücksichtigung der Heterogenität und Querschnittsdependenz, sondern auch der Interdependenz zwischen Vermögensmärkten ermöglicht. Die empirischen Ergebnisse einer trivariaten Vektorautoregression (VAR) mit realen Hauspreisen, realen Aktienkursen und realem persönlichen Pro-Kopf-Einkommen im Zeitraum von 1975 bis 2012 verdeutlichen die Existenz einer unidirektionalen Kausalität, die von beiden Vermögenspreisen zur Produktion verläuft. Dies bestätigt die Eigenschaft der Vermögenspreise als führende Indikatoren für die Realwirtschaft und dient zugleich als Beleg für den Mechanismus zur Übertragung von Vermögen und/oder Sicherheiten.

Emirmahmutoglu F., Balcilar M., Apergis N., Simo-Kengne B. D., Chang T. y Gupta R. Relación causal entre los precios de los activos y la producción en los Estados Unidos: evidencia de la prueba de causalidad de Granger en un panel de ámbito estatal, Regional Studies. En este artículo investigamos la relación causal entre los precios de los activos y la producción en los Estados de EE.UU. mediante un enfoque de bootstrap de causalidad de Granger en un panel que nos permite tener en cuenta no solo la heterogeneidad y la dependencia transversal sino también la interdependencia entre los mercados de activos. Los resultados empíricos de una autorregresión vectorial (VAR) trivariada que consta de precios reales de la vivienda, cotizaciones bursátiles reales y los ingresos personales reales per capita durante el periodo de 1975 a 2012 indican la existencia de una causalidad unidireccional que va de los precios de ambos activos a la producción. Esto confirma la propiedad de los precios de los activos como indicador líder para la economía real, pero también sirve como evidencia para el mecanismo de transmisión de riqueza y/o garantías.

• House prices
• Stock prices
• Output
• Granger causality

• 住宅价格
• 股票价格
• 产出
• 格兰杰因果关係

• Prix de l'immobilier
• Valeur des actions
• Production
• Causalité au sens de Granger

• Hauspreise
• Aktienkurse
• Produktion
• Granger-Kausalität

• Precios de la vivienda
• Cotizaciones bursátiles
• Producción
• Causalidad de Granger

JEL classifications:

• C33
• G10
• O18

Acknowledgement

The authors thank two anonymous referees for their many helpful comments that markedly improved the quality of the paper. However, any remaining errors are solely the authors’ alone.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental data

Supplemental data for this article can be accessed at http://dx.doi.org/10.1080/00343404.2015.1055462

Notes

1 Based on data from the US Department of Agriculture (USDA)-Economic Research Service, if the total agricultural production in a particular state as a percentage of the total agricultural production of the US is less than (greater than) 1%, the state is categorized as industrial (agricultural). See Appendix A in the Supplemental data online for agricultural versus industrial states.

2 The only study that can be considered related to this work is Chang et al. (2014). This study, based on an approach proposed by Kónya (2006), is a different panel causality test based on a seemingly unrelated regressions (SUR) estimator that yields a Wald test with country-specific bootstrap critical values, analysed bivariate causality between house prices and output for the nine provinces of South Africa. This test too does not require pre-testing for unit roots and co-integration apart from the lag structure. However, this approach does not provide a meta-analysis to help one conclude whether and which cross-sectional units drive the results for the entire panel.

3 As defined in the BEA regional accounts (see http://www.bea.gov/regional/definitions/):

Personal income is the income received by persons from participation in production, plus transfer receipts from government and business, plus government interest (which is treated like a transfer receipt). It is defined as the sum of wages and salaries, supplements to wages and salaries, proprietors’ income with inventory valuation and capital consumption adjustments, rental income of persons with capital consumption adjustment, personal dividend income, personal interest income, and personal current transfer receipts, less contributions for government social insurance. Because the personal income of an area represents the income that is received by, or on behalf of, all the persons who live in that area, and because the estimates of some components of personal income (wages and salaries, supplements to wages and salaries, and contributions for government social insurance) are made on a place-of-work basis, state personal income includes an adjustment for residence. The residence adjustment represents the net flow of compensation (less contributions for government social insurance) of interstate commuters.

Note that data on state-level GDP are available from the regional database of the BEA. But there is a break in 1997 in the way the GDP data is measured. In light of this, the BEA has a cautionary note (see http://www.bea.gov/regional/docs/product/):

There is a discontinuity in the GDP-by-state time series at 1997, where the data change from SIC [Standard Industrial Classification] industry definitions to NAICS [North American Industry Classification System] industry definitions. This discontinuity results from many sources. The NAICS-based statistics of GDP by state are consistent with U.S. gross domestic product (GDP) while the SIC-based statistics of GDP by state are consistent with U.S. gross domestic income (GDI). With the comprehensive revision of June 2014, the NAICS-based statistics of GDP by state incorporated significant improvements to more accurately portray the state economies. Two such improvements were recognizing research and development expenditures as capital and the capitalization of entertainment, literary, and other artistic originals. These improvements have not been incorporated in the SIC-based statistics. In addition, there are differences in source data and different estimation methodologies. This data discontinuity may affect both the levels and the growth rates of GDP by state. Users of GDP by state are strongly cautioned against appending the two data series in an attempt to construct a single time series for 1963 to 2013.

In light of this, the present authors, as in the existing literature, prefer the usage of personal income as a proxy for output at the state level.

4 In order to save space, see Pesaran and Yamagata (2008) for details of estimators and for Swamy's test.

5 Though few agricultural states, namely, AZ, FL and OR, display relatively low p-values, indicating causality running from housing prices in these states to stock price, while stock price is found to cause house price only in NE. For the industrial states, stock price only causes house price in WY, with no evidence of reverse causality from house price to stock price.

6 All the results of the meta-analysis based on the average Wald test statistic also continue to hold based on the Fisher statistic, as developed and used by Emirmahmutoglu and Kose (2011). Complete details of these results are available from the authors upon request. Furthermore, Appendix B in the Supplemental data online discusses results using real GDP per capita as the proxy of state-level output.

7 To account for structural breaks, Appendix C in the Supplemental data online provides results from the subsample analysis.

8 These results are inconsistent with the aggregate-level-based causality discussed in Appendix D in the Supplemental data online, hence justifying the importance of not deriving state-level inference from the aggregate level results.