DynamicConsumptionBehaviorEvidencefromJapanese

書式設定, 書式設定,第 2 ,第 3 ,第 4 ,第 5 ,*,Dynamic Consumption Behavior:Evidence from Japanese Household Panel Data,Yukinobu Kitamura,Hitotsubashi University,Institute of Economic Research,August 5, 2005,1,1. Plan of my talk,Family Income and Expenditure Survey 2001-2002,Dynamic Consumption Behavior,Durability of Consumption,Liqudity,Constraints,2,2 Durability of Consumption,Hayashi (1997, chapter5) shows the relationship between,expenditure,x,ijt,and consumption,c,ij.,Consumption is,a flow from accumulated expenditure in the past.,(1),where,i,is the agent,j,is consumption good or item,t,is,time,M,is a certain time dimension.,However, expenditures on individual goods,x,ijt,are ex-,pressed as the flow of current and future consumption,c,ijt,+,n,such that,(2),3,3 Liquidity Constraints,Households with debt. Although these households may not borrow as much as they wish, they may not have had faced with liquidity constraints in the past.,Households without debt and seemingly without liquidity constraints. They can borrow as much as they like but in fact they do not borrow.,Households without debt and seemingly with liquidity constraints. They may face with liquidity constraints at the moment.,Distinction between 1. and 2.3. can be made according to debt information at hand. Distinction between 2. and 3. requires some assumptions to draw lines according to annual income and net savings.,4,4 Empirical Model,Consumption function in empirical analysis is defined such that,(3),where,con,=log of real consumption,disp,=log of real disposable income,v,t,=monthly dummy.,The expenditure model is to convert consumption,con,into expenditure,x,.,(4),5,Empirical Results,AR(1) regression, coefficients of food, housing, traffic and telecommunication, recreation exceed 0.5. There seem to exist habit formation or sticky consumption. On the other hand, coefficients of furniture, clothes, medical expenditure, education are less than 0.2, habit formation does not exist for these items.,AR(4) regression indicates smaller but stable coefficients over four lags. Differences in coefficients across goods are smaller and coefficients do not drop sharply after some lags.,It implies that non durable consumption shows some durability.,6,Autoregression Model,7,Expenditure Behavior by Items,(1) consumption has durability. This is consistent with the fact that coefficients of own lag are negative.,(2) coefficients of disposable income are significant in many cases. Consumers seem to face the disposable income constraints.,8,Expenditure Behavior by Items,Panel A: Maximum Likelihood Method,9,Expenditure Behavior by Items,Panel B: GMM (one-step),10,Liquidity Constraints,(1) Households with debt (debtinc=1,debtass=1) and households without debt and with low annual income and net savings (debtinc=2,debtass=2) face the disposable income constraint.,(2) For those households, parameter values, statistical significance and implications remain, more or less, the same.,(3) Households without debt and with high annual income and net savings face the disposable income constraint in MLE.,11,Expenditure Behavior by Annual Income,Panel A: Maximum Likelihood Method,12,Expenditure Behavior by Annual Income,Panel B: GMM (one-step),13,Expenditure Behavior by Net Savings,Panel A: Maximum Likelihood Method,14,Expenditure Behavior by Net Savings,Panel B: GMM (one-step),15,Policy Implications,Sensitivity of expenditure to disposable income turns out to be significant for most cases.,This result implies, at least in the short run, policy variables such as taxes and social security contribution could affect consumption.,Provided the indebted households may face liquidity constraints, such policies as income tax reduction against the amount of mortgage, interest payment deduction and property tax reduction against the amount of mortgage might be used to relax such constraints.,16,