습관형성 가설과 가계 소비

Abstract

In this thesis, we analyzed the existence of habit formation in Korean households using Korean Labor and Income Panel Study (KLIPS). After consecutive failures of finding supporting empirical evidences of Life Cycle-Permanent Income Hypothesis(LC-PIH), time-nonseparable preference function has been introduced as one possible alternative. Unlike LC-PIH, past consumption has continual effects on current utility in this time-nonseparable preference function throughout habit formation. Past expenditure can either increase or decrease present utility, and we indicate the former as habit formation and the latter durability of consumption. To estimate the existence of habit formation in households' consumption behavior, we used the model developed by Dynan(2000). The estimated equation is based upon Euler's equation approach, and the same model was once estimated by Rhee(2001) with Korean Household Panel Studies for the period 1993-1998. Rhee(2001) concluded that the regression coefficients of the consumption growth rate on lagged consumption growth rate were positive, which, as a result, shows that Korean households exhibit strong habit persistence in consumption behavior. Meanwhile, we mainly analyzed general living expenditure costs excluding consumption on durable goods in this thesis and then, additional food consumption data were used in the purpose of more accurate comparison to former studies. Analyzing general living expenditure does not require any additional assumptions such as certain goods are perfectly separable in the preference function, which is commonly presumed when studying food consumption expenditure exclusively. Therefore, empirical analysis on general living expenditure will better reflect consumption behavior in reality. The KLIPS data used in this analysis were 6-year time series data for the period of 1998-2003 and have over 2,000 households cross-sectionally. Because of the measurement error usually found in panel data sets, income data as well as non-housing assets and lump-sum income dummy were used as instrumental variables when employing Random Effect Error Component 2-Stage Least Squares devised by Baltagi(1981) to obtain consistent estimator. Contrary to Rhee(2001)'s, we found out that the lagged changes in consumption is negatively correlated with the current consumption changes, which rather demonstrates that Korean households' preferences exhibit durability instead of habit persistence. This result hold true for both general living expenditure and food consumption analysis. That is, if the current level of consumption is high, decisions will be made to decrease the future level of consumption, as Korean households seem to gain constant utility from past consumption as well. This result was consistent even when a number of changes were made in the instrumental variable combination. There are several possible reasons to explain the difference in empirical findings between Rhee(2001)'s and this thesis. First, it could be resulted from the difference of sample size. KLIPS data has much higher rate of maintaining original households, and consequently, much larger sample was used in this thesis. Second, demographical differences in sample households possibly had significant effects on results. Finally, because Korean households experienced several economic shocks after 1998, actual changes in consumption behavior from habit formation to durability may have happened. In addition, due to a number of crucial economic shocks in 1998 and 2002, it is possible that liquidity constraint was binding for Korean households, which may have considerable effects in deriving the result of this thesis. In consequence, further in-depth studies should be conducted for two topics above.;본 연구는 한국노동패널조사(KLIPS: 1999~2004)를 이용하여 우리나라 가계의 최적화 문제에서 효용함수의 기간간 분리성이 성립하는지, 나아가 소비 행동에서 습관형성 혹은 내구성의 존재를 지지하는 증거를 찾을 수 있는지를 실증적으로 분석하였다. 이를 위해 KLIPS 자료의 생활비와 식비 및 식료품 소비 지출 데이터를 대상으로 실증분석을 실시하였다. 패널 자료의 측정오차 문제를 고려하여 분석방법으로는 Baltagi(1981)의 Random Effect Error Component 2-Stage Least Squares를 사용하였으며, 소득증가율, 금융소득증가율, 일괄지불소득 더미 등을 도구변수로 이용하였다. 특히 본 연구에서 사용된 KLIPS 자료는 조사 대상 기간이 1998년에서 2003년인 데이터로써 습관 형성에 대한 국내연구 중에서는 가장 최근의 자료를 이용하였다. 또한 내구재 소비를 제외한 생활비 데이터를 활용할 수 있기 때문에, 기존 연구에서 빈번하게 이루어졌던 ‘식료품 소비는 타 재화와 완벽하게 분리되어 효용을 발생시킨다’는 추가적인 가정을 하지 않고도 가계의 일반적인 소비 행동에 대한 분석을 하였다. 본 연구의 분석 결과는 다음과 같이 요약할 수 있다. 모든 추정식에서 일관적으로 소비증가율이 1계 시차값과 통계적으로 유의한 음의 상관관계를 갖는 것으로 나타났다. 습관형성에 관한 계수 추정치는 -0.34 ~ -0.13의 범위에 분포하고 있었으며, 따라서 우리나라 가계의 소비에서는 습관형성보다 내구성이 더욱 우세하게 나타나는 것으로 보인다. 이 연구 결과는 도구변수 조합에 여러 변화를 주었을 때에도 동일하게 나타나는 현상이다. 그러므로 가계는 과거 소비로부터 현재에도 지속적으로 효용을 얻으며, 과거 소비증가율이 높았다면 이번 기의 소비증가율을 감소시키는 선택을 한다고 볼 수 있다.