BIS Working Papers
The role of household debt heterogeneity on consumption: Evidence from Japanese household data
by Jouchi Nakajima
Monetary and Economic Department
JEL classification: D14, D81, D84, E21, G11
Keywords: household debt, marginal propensity to consume, precautionary saving motive
This publication is available on the BIS website (www.bis.org).
© Bank for International Settlements 2017. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.
ISSN 1020-0959 (print)
ISSN 1682-7678 (online)
The role of household debt heterogeneity on consumption: Evidence from Japanese household data
This paper estimates the impact of household debt on consumption behaviour using data from the Japanese Preference Parameters Study. Covering the 2005-13 period, the survey is the first of its kind for Japan. It features responses to forward-looking questions about key risks to income, shedding light on the motives for household savings behaviour. The analysis finds that household marginal propensities to consume (MPCs) were significantly higher for highly-indebted Japanese households than for those with little-to-no debt – a type of variation that is consistent with findings for other countries. The evidence points to a significant precautionary saving motive by Japanese households, with savers particularly concerned about (unlikely) future unemployment spells and longevity risks.
JEL classification: D14, D81, D84, E21, G11
Keywords: household debt, marginal propensity to consume, precautionary saving motive
Household debt has risen sharply around the globe in the past two decades, and that debt is seen as playing a more prominent role in economic decisions. The rapid declines in consumption by highly indebted households in the wake of the Global Financial Crisis (GFC) highlighted this role. And since the GFC, debt levels have continued to hit new highs. These experiences have naturally spurred considerable research into the relationship between household debt and consumption behaviour from both theoretical and empirical perspectives.
One strand of the literature has highlighted the significant macroeconomic impact of aggregate household debt on the economy. Mian and Sufi (2010) show that the growth in household leverage in the pre-GFC period explains a large fraction of the drop in consumption during the GFC (see also IMF (2017), Mian et al (2017), Mian and Sufi (2011, 2018)). Lombardi et al (2017) provide empirical evidence that even if household debt boosts the economy in the short run, it tends to impede growth in the long run. Drehmann and Juselius (2012) and Juselius and Drehmann (2015) find that debt service burdens negatively affect credit and expenditure growth, and can signal the likelihood of a future recession and financial crisis.
Another strand of literature delves into heterogeneous household debt burdens and their effects on consumption. Baker (2018) finds that highly indebted households are more sensitive to income changes because of borrowing and liquidity constraints. Kovacs et al (2017) use household-level data and show that highly leveraged households made larger cuts in their consumption following the crisis compared to average households, and this particularly the case for young households. The debt vulnerabilities generally vary with the type of debt instruments held by households. Floden et al (2017) for example find that highly indebted households with floating rate mortgages are more likely to cut their consumption growth by more in response to increases in interest rates than households with little debt or fixed rate mortgages. From another perspective, Kaplan et al (2018) develop the heterogeneous agent new Keynesian (HANK) models where households save in liquid and illiquid assets, suggesting that the household consumption is driven by more wealth effect than traditional intertemporal substitution effect. In such an environment, it is emphasised that the monetary policy effectiveness depends on fiscal response to the monetary expansion because of failure of Ricardian equivalence (see also Mian et al (2013), Kaplan et al (2014), La Cava et al (2016), Auclert (2017)).
This paper assesses the impact of Japanese household debt burdens on consumer behaviour using a panel dataset based on the Preference Parameters Survey (PPS). The paper is the first to find significant heterogeneity between marginal propensities to consume (MPCs) and debt burdens for Japanese households. One unique feature of this Japanese household survey, compared to previous surveys, is its compilation of responses to forward-looking questions. These data allow us to explore economic motives of Japanese savers.
We find empirical evidence that household MPCs are significantly higher for highly-indebted households than households with no debt. In the previous literature,precautionary savings due to liquidity and credit constraints have been highlighted (eg Mian et al (2013), Baker (2018)). The PPS dataset confirms the importance of the precautionary saving motive for Japanese households. In addition to funding consumption plans in the latter stages of life, Japanese savings behaviour reflects households' desire to self-insure against the unlikely event of a future unemployment spell and mortality risk.
The rest of the paper is organised as follows. Section 2 explains the econometric methodology to assess the relationship between household consumption patterns and the debt. Section 3 describe the main features of Japan's household-level data from the PPS. Section 4 presents estimation results from the panel-data estimation, and Section 5 provides concluding remarks.
2. The econometric methodology
The marginal propensity to consume (MPC) can be decomposed into income elasticity and average propensity to consume (APC); namely,
where C is consumption and Y is income; the left-hand side of the equation is the MPC, and the first and second terms on the right-hand side are the income elasticity and APC, respectively. In the following analysis, the income elasticity is estimated with a panel regression model, and the APC is calculated as the sample average within a respective group of households. This approach enables us to assess whether differences in the MPC across groups is attributable to the income elasticity or the APC.3 The analysis focuses on the heterogeneity of the consumption pattern across households with different levels of debt burden.
To estimate the income elasticity and its link to debt burdens, we consider the following panel regression model:
where Aln(cit) is the annual change in expenditure of household t in the year t, Aln(yit) is the annual change in income, DRit is a variable of debt ratio, Xit is a vector of household characteristics, and at is the household fixed effect.4 The coefficient ^ measures the income elasticity for a household without any debt. The coefficient p2 measures an effect of holding debt on the income elasticity. For a household with a certain amount of debt with the debt ratio DR, the income elasticity is p1 + p2DR.
The analysis takes account of the possible asymmetry in the consumption patterns between positive and negative income changes. As a baseline model, the following regression is estimated:
This specification implies that the income elasticity is p4 + P2DRit for households facing a positive income change in the current year and p3 + P4DRit for those facing a negative income change. The coefficients p2 and p4 measure the effects of household debt on the income elasticity in the cases of positive and negative income changes, respectively.
Further, the analysis addresses the saving motives as one possible factor to account for the relationship between debt and consumption. To empirically test its relevance, the following regression is estimated:
where Zit is a dummy variable indicating a saving motive for household i in the year t. The coefficients p2 and p6 reflect the baseline effect of the saving motive on the income elasticity, and p4 and p8 measure an additional effect of holding debt on the income elasticity. For example, when a household with a debt ratio DR faces a positive income change, the income elasticity is represented by p1 + p2Zit + DT? x (P3 + P4Zit). If the household does not have a saving motive (ie Zit = 0), the income elasticity is p1 + p3DR. If it has a saving motive (ie Zit = 1), the income elasticity becomes p1 + p2 + (J33 + p4)DR. A statistical test on the coefficient p4 can address whether the heterogeneity of the income elasticity across different level of indebtedness arises from the household's saving motive.
3.1 Household survey
The analysis exploits the Japanese household survey from the Preferences Parameters Study (PPS) conducted by the Institute of Social and Economic Research at Osaka University. This PPS dataset includes information on households' characteristics (eg age, employment status, residence region), income, expenditure, balance sheet, and future expectations.5 The survey was conducted annually from 2003 to 2013. The sample of households were randomly selected from the Basic Residents Registration System to provide a representative sample of Japanese households.6 The dataset is an unbalanced panel. Households that did not reply to the survey in a given year were dropped from the survey in subsequent years. New households were added in 2004, 2006, and 2009. The dataset includes an average of 4,000 samples for each year.
One advantage of this survey for the current analysis is that the dataset includes answer to questions about each households' expectations about the future. In particular, households' responses to questions about expected future employment and old age situations. These responses allow us to assess the importance of the precautionary saving motives and the role of debt for household consumption behaviour.
3.2 Selected variables and summary statistics
Details about the variables selected for the analysis are described in this section. Missing questionnaires in 2003 and 2004 restrict the useable survey data to be from 2005 to 2013 for the panel regression analysis.8 Table 1 reports descriptive statistics of the variables described in the following.
The dependent variable in the regression is the annual change in expenditures. The survey asks about the annual change in a household's total expenditure and offers categorical choices for responses: starting from "decrease by more than 9%", "decrease by 5%-7%", "decrease by 3%-5%", to "increase by 7%-9%" and "increase by more than 9%." For the estimation, the mid-point of each category range is assigned a numeric variable, following Kubota and Fukushige (2016). The first and last open-ended categories are replaced with -11% and 11%, respectively.9 Over the sample, the average annual expenditure growth is 1.35%. The cross-sectional averages over time exhibit a similar trend to that of national consumption growth.
The main explanatory variable in the regression is the annual change in income. The questionnaire offers the same categorical choices for this variable as for the annual change in expenditures. The same numeric assignment is applied to the income change. Over the sample, the average annual income growth is -0.87%, and cross-sectional averages over time are similar to the trend in national earnings growth.
The key explanatory variable in the regression is the cross-term of the annual income change with the debt ratio (DR). In the regression analysis, three debt
ratios are examined: the ratio of debt over assets (D/A), over financial assets (D/FA), and over income (D/l). In the survey, the debt (D) is defined as total housing loans, car loans, and any other instalment payments on which the household has to pay interest charges.10 The assets (A) are the appraised value of all housing, property, and financial assets (FA) such as savings, stocks, and insurance owned by the household. The income (l) is annual earnings before taxes plus bonuses of the entire household.11 The sample averages of D/A, D/FA, and D/l are 0.31, 2.04, and 1.51, respectively.12 Note that 46 percent of the sample households report having no debt.
The model regresses the annual change in income onto the change in expenditure to measure the elasticity of consumption. One of caveats in this analysis is that we cannot identify whether income changes are permanent or temporary. The samples in the time series dimension are not adequate, and the dataset is unbalanced as households are invited to answer the survey only for, on average, four consecutive
In addition to the baseline specification, a robustness check examines unexpected changes in income and expenditure, using the answers to the questions about the expected changes in the next-year income and expenditure. This specification can assess a household's reaction to an unexpected income change and its relation to the household's debt burden. The responses for the expected income and expenditure changes are the same categorical choices as for the realized annual changes in income and expenditure. The same numeric assignment is applied for the expected changes.
Two important questions associated with precautionary saving motives address concerns about future unemployment spells (CON_UEMP) and old age (CON_OLD):
Q. What is the possibility that you or your spouse will be unemployed within the next two years? Choose one of the following: 1. Strong possibility, 2. Some possibility, 3. Little possibility, 4. Don't know.
Q. I have anxieties about my 'life after I am 65 years old.' Answer on a scale from 1 to 5, where "1" means it is particularly true and "5" means that it doesn't hold true at all. In the estimation, these categorical choices are converted to dummy variables that take a value of one or zero. Specifically, the value for CON_UEMP is one if the answer to the question is "1", and zero otherwise. The value for CON_OLD is one if the answer is either "1" or "2", and zero otherwise. As shown in Table 1, these concerns are notably common in Japanese households; 22% and 54% of the samples have concerns about unemployment and old age, respectively.
The control variables are the following household characteristics that may possibly affect the annual change in expenditure:
The survey asks about the number of household members. A change in household composition from the previous year is calculated as the control variable. When the number of household members increases (decreases), it is expected to increase (decrease) the annual consumption expenditure (eg Kubota and Fukushige (2016)).
The survey includes a question about job categories of the household members, such as farming, construction, and financial operations. A dummy variable is created that takes a value of one if the job categories have changed from previous year, and zero otherwise. Changing a job may affect the expenditure beyond the income change.
The information about households' residential area is used as a control variable. The survey identifies the household's residential area by four city size categories: cabinet-designated city (a city with population greater than 500,000), population over 100,000, population less than 100,000, and town or village. A dummy variable is created that takes a value of one (minus one) if the household moves to a larger (smaller) city in the current year, and zero otherwise.
4. Empirical results
4.1. Baseline results
Table 2 shows estimation results for baseline regression model (1), with each specification examining a different debt ratio: debt over assets (D/A), over financial assets (D/FA), and over income (D/l). The income elasticity of consumption measured by the coefficient of the annual income change is around 0.24 to 0.28, with little difference between the positive and negative income changes.
The coefficient of the cross-term of the annual change in income with the debt ratio (DR) is not significant for the positive income change, but it is significantly positive for the negative income change. This result implies that when the households face a decline in income, those with a higher level of debt tend to save more than Ito et al (2017) use similar variables to assess the precautionary saving motive to analyse households' portfolio selection. See also Miura and Higashi (2017). those with no debt or a lower level of debt. This relation holds for all specifications with different debt ratios reported in Table 2.
To compute the MPC for different groups of households by the debt ratio (debt over assets), the APC is calculated for households with DR = 0 ("Low" debt ratio) and DR > H ("High"). The latter group is defined as households holding debts with a debt ratio exceeding the threshold H. Specifically, H = 0.36 is used here, which corresponds to the upper quartile of the distribution of DR across the entire sample. Graph 1 illustrates the estimate of APCs, which are about 0.75 and 0.84 for the groups of DR = 0 and DR > H, respectively. It shows the slightly higher APC for the highly- indebted households, although little difference exists between the cases of facing positive and negative income changes.
Graph 1 also shows the resulting MPCs for the two groups of household in the cases of positive and negative income changes. The estimate indicates that the MPC is considerably higher for households with debt when they face a negative income change. The MPC for no-debt households is around 0.20, as is the MPC for highly-indebted households when they gain income, while it increases to 0.28 when the highly-indebted households face the negative income change. Note that the degree of income elasticity, APC and MPC obtained here are consistent with the estimate by the Bank of Japan (2016).
What underlies the households' consumption behaviour? The next regression analysis examines the precautionary saving hypothesis.
4.2 Precautionary saving motive
Theoretically, precautionary saving motives are modelled as arising when constant relative risk aversion utility and certain liquidity constraints are assumed, as shown by Deaton (1991) and Carroll and Samwick (1997).15 A consumer who faces uncertainty on income shocks holds a positive amount of wealth to have a buffer stock of savings. A wide range of studies provide empirical support for consumers' precautionary saving behaviour (eg Dynan (1993), Horioka and Watanabe (1997), Carroll and Samwick (1998), Lusardi (1998), Engen and Gruber (2001), Cagetti (2003), Hochguertel (2003), Fisher and Montalto (2010)). A close look at precautionary motives not only yields a better understanding of individual consumers' behaviour, but also provides macroeconomic and policy implications (Zeldes (1989)). For example, Caballero (1990) argues that consumers' precautionary saving motives explain several facts about aggregated consumption, such as its excessive smoothness and excessive sensitivity to anticipated income changes. The current paper sheds light on the importance of the debt burden to consider different MPC across households and their precautionary saving when they face a decline in income.
Table 3 reports estimation results for panel regression model (2), with the dummy variable for concerns about (a) unemployment and (b) old age. In this regression, the debt ratio over asset is used for the cross-term of the annual change in income. The results show that the coefficient of the cross-term with the dummy variable for concerns is significantly positive only for the case of a negative income change, while coefficients of other cross-terms are not significant. This outcome supports the precautionary saving hypothesis, as the concern about unemployment risk increases the income elasticity through the cross-term of the annual change in income with the debt ratio.
Graph 2 presents the estimated income elasticity, APC and MPC for households with precautionary saving motives. The "High" group of debt ratio is the households with DR>H and with concerns about unemployment risk (CON_UEMP=1). The resulting MPC is higher for highly-indebted households than no-debt households for the case of a negative income change. In addition, the MPC is lower for highly- indebted households for the case of a positive income change. This result implies that the highly-indebted households tend to save more when they have concerns about income risk due to future unemployment and uncertainty about resources in old age.
The literature has emphasised the possibility of a precautionary saving motive that arises from liquidity and credit constraints linked to debt burdens (eg Mian et al (2013), Baker (2018)). Highly-indebted households are more likely to be bound by liquidity and credit constraints. In such circumstances, households have an incentive to save more as incomes increase than those households with low debt levels, thereby relaxing the constraints. In other words, highly-indebted households would have lower MPCs than those with low debt levels. When they face a decline in income, they cut their spending to have enough buffer from the constraint, which makes their MPC higher. The Japanese dataset confirms this household behaviour and reveals that the concerns about the latter stages of life and unemployment risks are key determinants of the precautionary saving motive.16
Table 4 reports estimation results of the panel regression model (1) for different age cohorts. Significant consumption differences among the cohorts are found. For a decline in income, consumption choices by cohorts of (b) ages 36-50 and (c) ages 51-65 are sensitive to debt levels (the coefficient on the cross-term of the income change and debt is significantly positive). For an increase in income, the cohort of (b) ages 36-50 is also significant. These relationships are not found for the cohorts of (a) ages 20-34 and (d) ages 65-80.
The results indicate that debt burdens significantly influences consumption behaviour of pre-retirement middle-aged households, but not of younger households. These differences are consistent with housing debt playing an important role in consumption behaviour. In the earlier stages of life, few Japanese households own their housing. Their major component of debt is credit card loans. For middle- aged households (from the late 30s to the early 60s), Japanese households typically buy houses and take on a high level of debt. This debt overhang appears to make their saving behaviour more sensitive to income changes. In particular, highly- indebted middle-aged households up to their 40s tend to save much more when income increases. This is consistent with evidence that Japanese households are likely to pre-pay mortgage debt (Kobayashi (2012)). After the standard retirement age of 65, most households have paid down their mortgage debt; along with a relatively lower level of debt and pensions being a major source of income, consumption behaviour appears less sensitive to debt.
Using another Japanese household survey conducted in 1994, Horioka and Watanabe (1997) show the existence of saving motives due to retirement and illness. They also show that Japanese households tend to save for debt service related to housing debt. Murata (2003) provides empirical evidence that Japanese households tend to have precautionary savings if they have concerns about the uncertainty of public pension benefits, particularly when they cannot receive money transfers from their parents. The current result adds the new evidence to this area of studies on Japanese households' saving behaviours.
To check the robustness of the estimation results, the panel regression model is estimated with different sets of variables. First, Table 5(1) shows the regression results with the expected change in the next-year income additionally included as a control variable. The coefficients of the expected income change are significantly positive (negative) for positive (negative) expected change, which implies households' consumption smoothing behaviour. The coefficient for the cross-term of the current income change with the debt ratio is significantly positive for the case of facing a negative income change. This finding is still valid for the specification with the annual changes in expenditure and income replaced by unexpected changes, as shown in Table 5(2).
Table 6(1) reports another regression with a level of debt for the cross-term, instead of the debt ratio, which shows the same significance for the cross-term of the highly-indebted households. Table 6(2) examines the annual change in income relative to national income change to control for the influence of the national business cycle. Further, the number of household members and their ages, their job classifications, the geographic locations of their residences and sizes of the city in which they live are treated as additional control variables. The results confirm the robustness of the reported findings.
5. Concluding remarks
This paper assesses the impact of Japanese household debt burdens on consumer demand using the household data for Japan. The analysis reveals that MPCs are higher for households with higher levels of debt, especially for negative income shocks. The estimation results indicate that precautionary saving motives play an important role in making highly-indebted households more sensitive to income decline due to concerns about future unemployment risks and about having adequate savings for the latter stages of life.
The results have some policy implications. A policy that decreases future uncertainty about income during spells of unemployment and about adequate income during old age would be beneficial for highly-indebted households. For example, Engen and Gruber (2001) demonstrate the benefit of unemployment insurance for consumption behaviour. With increasing life expectancy, life after retirement becomes one of the more serious concerns. Mortality risk implies that well funded pension and health insurance systems can influence households' consumption patterns. Guiso et al (1992) point out that uncertainty about health is an important factor for precautionary saving.
Murata (2003) finds that Japanese households tend to have precautionary saving motives due to uncertainty about the public pension scheme, which implies that a more sustainable pension restructuring could be key when addressing the challenges posed by a progressively aging society. Japanese elderly households heavily depend on public pension as a dominant source of income after retirement. The pension restructuring has been repeatedly discussed including an increase in the normal pension age (eg Fukawa (2006)), while timing of the reform is typically unknown. This policy uncertainty can increase uncertainty on the main source of income after retirement, which can affect the households' precautionary saving behaviour. Kitao (2018) points out an adverse effect of delaying the reform in Japan using a general equilibrium life-cycle model. Assessing changes in households' saving behaviour responding to the pension reform with an age-cohort analysis would be an important future work.
One caveat in the PPS dataset is that there is no information about the debt service burden which each household faces. However, if the wealth effect is more dominant than the intertemporal substitution effect in response to monetary policy as eg Kaplan et al (2018) show, the household debt heterogeneity on MPCs found in this paper suggests the importance of the household balance sheet information in considering the monetary policy effectiveness. A quantitative assessment of both intertemporal-substitution and wealth effects with the debt heterogeneity in the household sector remains as a relevant future work.
The PPS dataset includes the survey not only of households in Japan but also households in the United States, China, and India. Although the latter datasets are relatively smaller and more limited than the Japanese one, comparing the role of household debt and its effect on consumption expenditures across the countries could shed further light on the issues raised in this paper. Several Asian countries have faced rising debt (eg BIS (2009) and ADB (2017)). However, for China and India, assessment of the role of debt with large-scale household panel data is limited. Chamon et al (2013) use Chinese data and point out an increasing saving rate in China's urban households; they show that rising income uncertainty leads to a significant increase in households' saving.17 With a careful examination of data quality, the Preferences Parameters Study has the potential to provide new insights for those Asian countries. Such investigation is left as a future work.
Aron, J, J V Duca, J Muellbauer, K Murata and A Murphy (2012): "Credit, housing collateral, and consumption: evidence from Japan, the UK, and the US", Review of Income and Wealth, vol 58(3), pp 397-423.
Auclert, A (2017): "Monetary policy and the redistribution channel", NBER Working Paper, no 23451.
Asian Development Bank (2017): Asian Economic Integration Report 2017
Baker, S R (2018): "Debt and the response to household income shocks: validation and application of linked financial account data", Journal of Political Economy, forthcoming.
Bank for International Settlements (2009): "Household debt: implications for monetary policy and financial stability", BIS Papers, no 46.
Bank of Japan (2016): Outlook for Economic Activity and Prices, October.
Caballero, R J (1990): "Consumption puzzles and precautionary savings", Journal of Monetary Economics, vol 25(1), pp 113-36.
Cagetti, M (2003): "Wealth accumulation over the life cycle and precautionary savings", Journal of Business and Economic Statistics, vol 21(3), pp 339-53.
Carroll, C D and A A Samwick (1997): "The nature of precautionary wealth", Journal of Monetary Economics, vol 40(1), pp 4-71.
-------- (1998): "How important is precautionary saving?", Review of Economics and
Statistics, vol 80(3), pp 410-19.
Chamon, M, K Liu and E Prasad (2013): "Income uncertainty and household savings in China", Journal of Development Economics, vol 105, pp 164-77.
Deaton, A (1991): "Saving and liquidity constraints", Econometrica, vol 59(5), pp 122148.
Debelle, G (2004): "Macroeconomic implications of rising household debt", BIS Quarterly Review, March 2004.
Drehmann, M and M Juselius (2012): "Do debt service costs affect macroeconomic and financial stability?", BIS Quarterly Review, September 2012.
Dynan, K E (1993): "How prudent are consumers?", Journal of Political Economy, vol 101(6), pp 1104-13.
-------- (2012): "Is a household debt overhang holding back consumption?", Brookings
Papers on Economic Activity, vol 2012(1), pp 299-362.
Engen, E M and J Gruber (2001): "Unemployment insurance and precautionary saving", Journal of Monetary Economics, vol 47(3), pp 545-79.
Fisher, P J and C P Montalto (2010): "Effect of saving motives and horizon on saving behaviors", Journal of Economic Psychology, vol 31(1), pp 92-105.
Floden, M, M Kilstrom, J Sigurdsson and R Vestman (2017): "Household debt and monetary policy: revealing the cash-flow channel", CEPR Discussion Paper, no DP12270.
Friedman, M (1957). A theory of the consumption function, Princeton: Princeton University Press.
Fukawa, T (2006): "Sustainable structure of the Japanese public pension system viewed from a Germany-Japan comparison", Japanese Journal of Social Security Policy, vol 6(1), pp 131-43.
Guiso, L, T Jappelli and D Terlizzese (1992): "Earnings uncertainty and precautionary saving", Journal of Monetary Economics, vol 30(2), pp 307-37.
Hochguertel, S (2003): "Precautionary motives and portfolio decisions", Journal of Applied Econometrics, vol 18(1), pp 61-77.
Horioka, C Y and W Watanabe (1997): "Why do people save? A micro-analysis of motives for household saving in Japan", Economic Journal, vol 107(442), pp 537-52.
International Monetary Fund (2017): "Household debt and financial stability", Global Financial Stability Report, October 2017.
Ito, Y, Y Takizuka and S Fujiwara (2017): "Portfolio selection by households: an empirical analysis using dynamic panel data models", Bank of Japan Working Paper Series, no 17-E-6.
Juselius, M and M Drehmann (2015): "Leverage dynamics and the real burden of debt", BIS Working Papers, no 501.
Kaplan, G, B Moll and G L Violante (2018): "Monetary policy according to HANK", American Economic Review, vol 108(3), pp 697-743.
Kaplan, G, G L Violante and J Weidner (2014): "The wealthy hand-to-mouth", Brookings Papers on Economic Activity, Spring, pp 77-153.
Keynes, J M (1936). The general theory of employment, interest and money. London: MacMillan.
King, M (1994): "Debt deflation: theory and evidence", European Economic Review, vol 38(3-4), pp 419-45.
Kitao, S (2018): "Policy uncertainty and cost of delaying reform: the case of aging Japan", Review of Economic Dynamics, vol 27, pp 81-100.
Kobayashi, M (2012): "Asian mortgage markets", 2012 EMF HYPOSTAT, pp 20-7.
Kovacs, A, M Rostom and P Bunn (2017): "Debt and consumption in the United Kingdom during the crisis", Working Paper.
Kubota K and M Fukushige (2016): "Rational consumers", International Economic Review, vol 57(1), pp 231-54.
La Cava, G, H Hughson and G Kaplan (2016): "The household cash flow channel of monetary policy", Reserve Bank of Australia Research Discussion Papers, no 2016-12.
Lombardi, M J, M S Mohanty and I Shim (2017): "The real effects of household debt in the short and long run", BIS Working Papers, no 607.
Lusardi, A (1998): "On the importance of the precautionary saving motive", American Economic Review, vol 88(2), pp 449-53.
Mian, A, K Rao and A Sufi (2013): "Household balance sheets, consumption, and the economic slump", Quarterly Journal of Economics, vol 128(4), pp 1687-726.
Mian, A and A Sufi (2010): "Household leverage and the recession of 2007-09", IMF Economic Review, vol 58(1), pp 74-117.
-------- (2011): "House prices, home equity-based borrowing, and the US household
leverage crisis", American Economic Review, vol 101(5), pp 2132-56.
-------- (2018): "Finance and business cycles: the credit-driven household demand
channel", NBER Working Paper Series, no 24322.
Mian, A, A Sufi and E Verner (2017): "Household debt and business cycles worldwide", Quarterly Journal of Economics, vol 132(4), pp 1755-817.
Mishkin, F S (1978): "The household balance sheet and the Great Depression", Journal of Economic History, vol 38(4), pp 918-37.
Miura K and M Higashi (2017): "The recent increase in dual-income households and its impact on consumption expenditure", Bank of Japan Review, no 2017-E-7.
Murata, K (2003): "Precautionary savings and income uncertainty: evidence from Japanese micro data", Monetary and Economic Studies, vol 22(3), pp 23-58.
Rungcharoenkitkul, P (2011): "Wealth effects and consumption in Thailand", Bank of Thailand Discussion Paper, no 2011-01.