Household heterogeneity and consumption dynamics in the presence of borrowing and liquidity constraints

ABSTRACT

I study the effects of borrowing and liquidity constraints on the response of consumption to anticipated income changes. Using the PSID over 1999–2013, I find that the well-documented strong excess sensitivity of consumption to income of highly constrained households can be explained by episodes of income increases. In addition, I look into the heterogeneity of households without debt, a group that has been largely disregarded by the literature. My fixed-effects estimates show that only those without debt tend to increase their saving in response to anticipated income declines, irrespective of the amount of liquid assets held.

KEYWORDS:

• Consumption
• income
• highly constrained households
• household heterogeneity

JEL CLASSIFICATION:

• C81
• D12
• E21

Acknowledgements

The author thanks Gert Peersman, Sebastian Rüth, and participants at the Macro Research Group meeting, the 21st T2M 2017 Conference, Lisboa, Portugal, and the 22nd SMYE 2017 Conference, Halle (Saale), Germany, for their comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

I focus on anticipated income changes and not on unanticipated income shocks. The former has the advantage of not making any assumption about the income process, with income shocks and the error term being modelled jointly. Jappelli and Pistaferri (2010) provide a comprehensive overview about the two different concepts. Nevertheless, in ongoing research, I exploit my dataset against the backdrop of unexpected income shocks.

² There have been, however, more recent studies on the consumption response to unanticipated transitory positive and negative income shocks, such as Bunn et al. (2017) and Christelis et al. (2017).

³ Own tabulations of the PSID dataset available upon request.

⁴ To keep the analysis parsimonious, I proxy highly constrained households with: the first quintile of the LAR with the top two DSR quintiles (LAR Q1 $\mathrm{\&}$ DSR Q4-Q5); and the LAR < 2.5 $\mathrm{\%}$ with the top two DSR quintiles (LAR <2.5$\mathrm{\%}$ $\mathrm{\&}$ DSR Q4-Q5).

Additional information

Funding

This work was supported by the UGent Special Research Fund (BOF).