非線性消費行為與生命循環／恆常所得假說

Nonlinear Consumption Behavior and Its Implaction on the LCH/PIH

本文研究台灣非耐久財消費是否具有非線性，及其對生命循環／恆常所得假說檢定的影響。實證結果發現，台灣的非耐久財消費具有顯著的非線性門檻效果，顯示以往在線性模型下的推論有設定錯誤的可能。因此，本文改以一消費門檻模型來描述消費成長的行為，並重新檢定生命循環／恆常所得假說。相較於國內過去的實證結果普遍拒絕生命循環／恆常所得假說，本文結果顯示，生命循環／恆常所得假說是否得到支持，受景氣循環的影響；當經濟景氣較佳時（如，較高的所得成長率、貨幣供給成長率或股票報酬率），非耐久財消費對當期所得變動的反應並不顯著，符合生命循環／恆常所得假說的預測；反之，在景氣較差時，消費對所得變動有過度敏感的現象；接著，我們再針對生命循環／恆常所得假說不成立的期間，檢定消費對可預測所得的變動是否具有對稱反應，以探究造成不成立的原因是短視或流動性限制。本文發現，流動性限制存在較能解釋消費過度敏感的現象，進一步釐清國內現有文獻認為，流動性限制無法解釋生命循環／恆常所得假說不成立的推論。 This paper investigates the nonlinearity of consumption behavior and its implication on the life cycle hypothesis/permanent income hypothesis (hereafter LCH/PIH). Using the nondurable consumption data in Taiwan, we find a significant threshold effect on the consumption behavior, which means the previous studies with linear regression models might potentially be subject to the misspecification problem. A threshold regression model that better captures the consumption process is thus utilized, and the tests on the LCH/PIH is re-conducted. Contrasting with the existing empirical results that almost universally reject the LCH/PIH, we find that the validity of the LCH/PIH depends crucially on different phases of the business cycle. Excess sensitivity of consumption to predictable income can only be found during the phases of recession, revealing the fact that the consumption behavior matches what the LCH/PIH predicts only during the periods of economic expansion. To further explore the cause of the empirical failure of the LCH/PIH, we extract data from the periods when there is excess sensitivity of consumption. Our finding that consumption responds asymmetrically to the predictable income leads to the conclusion that the rejection of the LCH/PIH results from the existence of the liquidity constraints, rather than myopia.

This paper investigates the nonlinearity of consumption behavior and its implication on the life cycle hypothesis/permanent income hypothesis (hereafter LCH/PIH). Using the nondurable consumption data in Taiwan, we find a significant threshold effect on the consumption behavior, which means the previous studies with linear regression models might potentially be subject to the misspecification problem. A threshold regression model that better captures the consumption process is thus utilized, and the tests on the LCH/PIH is re-conducted. Contrasting with the existing empirical results that almost universally reject the LCH/PIH, we find that the validity of the LCH/PIH depends crucially on different phases of the business cycle. Excess sensitivity of consumption to predictable income can only be found during the phases of recession, revealing the fact that the consumption behavior matches what the LCH/PIH predicts only during the periods of economic expansion. To further explore the cause of the empirical failure of the LCH/PIH, we extract data from the periods when there is excess sensitivity of consumption. Our finding that consumption responds asymmetrically to the predictable income leads to the conclusion that the rejection of the LCH/PIH results from the existence of the liquidity constraints, rather than myopia.