Household debt, student loan forgiveness, and human capital investment: a neo-Kaleckian approach

Abstract

This paper aims to analyze the sustainability of student debt in the US. For this purpose, I build a neo-Kaleckian model in which households can borrow to either consume or invest in human capital. Next, I calibrate the model using US data to simulate the economic effects of specific policies such as student loan forgiveness. To my knowledge, this is the first study that considers household borrowing for two different purposes, consumption and human capital accumulation, in a demand-led macro-modeling framework. The main findings are that (i) household debt is sustainable in the long run (i.e., the debt servicing is compatible with the long-term economic growth) for a consumption level greater than 90% of household income; (ii) new borrowing boosts short-term economic activity while having ambiguous long-term effects because of its outcomes to household indebtedness and debt servicing; and (iii) student loan cancelation has only short-run economic effects, whereas reducing loan interest rates and changing the eligibility criterion for student loan forgiveness result in long-term effects.

Keywords:

• Capacity utilization
• household debt
• human capital
• post-keynesian economics
• student loans

JEL CLASSIFICATION:

• E11
• E12
• E22
• E24

Acknowledgments

Preliminary versions of this paper were presented at the 12th Annual PKES PhD Student Conference, May 2021, and the YSI Keynesian Economics Debate Group (KEDG) 2nd edition, October 2020. I thank Mark Setterfield, Duncan Foley, Gilberto Lima, Willi Semmler, Rafael Wildauer, Paulo Lins, conference participants, and one anonymous referee of this journal for their helpful comments and discussions. Any remaining errors are my own.

Notes

1 Data from the Federal Student Loan Portfolio - US Department of Education.

2 Data from the New York Fed Consumer Credit Panel/Equifax.

3 As I explain later in Appendix B, I calibrate human capital investment ($I_H$) using the Personal Consumption Expenditures: Education series, from the Bureau of Economic Analysis. Considering that, in the long-term dynamics, the growth rate of the human capital stock is a function of $I_H,$ the human capital variable ($H$) is measured in monetary terms, as the accumulation of previous human capital investment, similarly to the physical capital variable.

4 This model assumes a quantity adjustment in the short run, as explained in Section 3.3. Briefly, the goods market clears through an adjustment in the physical capital capacity utilization ($u$) so that the actual output level is $Y=\nu uK.$ The employment level ($L$) results from the labor required to produce $Y.$ From Equation (1), it implies $L=\frac{Y}{a(H/N)}.$

5 Dutt and Veneziani (2019) represent the labor productivity growth rate as given by ${\widehat{A}}_i={\tau }_0+{\tau }_1\frac{A_{i}H}{K},$ with $i=L,H$ standing for low- and high-skilled workers, respectively, and $H$ is the number of high-skilled workers. Another alternative is Neto and Ribeiro (2019), who also consider two types of labor in a neo-Kaleckian model, unskilled and skilled, with labor productivity being affected in the medium run by technological progress.

6 Dutt (2006) assumes that the desired level of borrowing is given by $\beta [(1-\pi )Y-iD]$ and, for simplicity, the amount of borrowing is equal to the desired level. Dutt develops an alternative approach in another paper (Dutt 2005), in which households target a desired level of debt. Similarly, in this model, I assume that Equation (2) represents both households’ desired and actual borrowing.

7 To understand this result, let $w$ be the wage per effective unit of labor, which also represents the incremental income of one additional unit of human capital. Assuming that from period $t=1$ onwards $w$ is taken as constant, the present value ($PV$) of the cash flow per unit of human capital investment, for a discount rate $r$ and a constant employment rate $e=1,$ is:

$PV={\sum }_{t=1}^{\infty }\mathrm{ }[w(\frac{1}{1+r}{)}^t]=\frac{w}{r}=\frac{(1-\pi )a}{r}$(5).

Since the wage share equals $(1-\pi )=\frac{\mathit{wLH}}{Y},$ using Equation (1) and $N=1$ implies $w=(1-\pi )a.$ Therefore, the present value of the cash flow is a function of the wage share.

8 For demand-led macroeconomic models that analyze income-driven repayment plans for student loans, see Serra and Lima (2018) and Carvalho, Lima, and Serra (2017).

9 Using Equation (1), the profit rate is given by $r=\frac{rK}{Y}\frac{Y}{K}=\pi \nu u.$

10 Equation 9 is obtained after dividing Equation (8) by $K.$ The left-hand side results in $Y/K=\nu u,$ where $u$ is the utilized fraction of the physical capital stock.

11 In the absence of this condition, with relatively higher responsiveness of investment than savings to excess demand, increases in capacity utilization would feedback into aggregate demand, raising the level of activity again and moving the economy further away from the equilibrium (Blecker and Setterfield 2019, 285). This dynamic process is verified, for instance, in the Harrodian formulation of the investment function, which considers an accelerator principle (Harrod 1939; Skott 2010). Some alternatives to tame the instability in those cases include an autonomous non-capacity creating component of aggregate demand, as in the Sraffian Supermultiplier literature (Freitas and Serrano 2015), or an offsetting monetary policy to the dynamics (Duménil and Lévy 1999).

12 Equations (10) and (11) show that increasing the autonomous component and responsiveness of new borrowing to changes in labor income, in this order, contribute to economic activity. On the other hand, a higher responsiveness to changes in the debt service reduces new borrowing and, consequently, aggregate demand (Equation (12)).

13 From Equation (21), $\partial h^{\mathrm{*}}/\partial \mathrm{\hspace{0.17em}\theta \hspace{0.17em}}$>0 if worker households’ new borrowing is positive.

14 From Equation (2), $(D_S/Y{)}_{\mathit{stable}}^{\mathrm{*}}$ represents a fraction $\theta =0.0740$ of $(D/Y{)}_{\mathit{stable}}^{\mathrm{*}}.$

15 Data from the International Monetary Fund.

16 To understand Table 3, assume, for illustration, that households consume 80% of their income (i.e., devote 20% of it to the debt service). In this case, the maximum debt service they can afford corresponds to a debt ratio of ${\delta }_{\max ,\varphi =20\mathrm{\%}}=0.2621.$

17 In the fourth quarter of 2019, the student debt balance-to-GDP was 6.9%. Thus, the simulation represents the cancellation of approximately 14.4% of the outstanding balance in the US. Noteworthy here is that only 7.9% of total outstanding student debt are related to private loans (Amir, Teslow, and Borders 2020) so that the proposed cancellation of government-owned student loans is feasible.

18 Fullwiler et al. (2018) present 10-year forecasts of debt cancellation using two models: the Moody’s and the Fair model. Although they both have Keynesian theoretical foundation in the short run, the former model features classical results in the long run, whereas the latter keeps the short-term assumptions.

19 Considering that student debt represents 9.4% of the total outstanding balance of households, this reduction of 0.1% represents a decrease of 1.06% of the student loan interest rate.

20 I follow Lavoie, Rodriguez, and Seccareccia (2004) in considering one lag for capacity utilization.

Additional information

Funding

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001.

Notes on contributors

Gustavo Pereira Serra

Gustavo Pereira Serra is Assistant Professor at Department of Economics, São Paulo State University (UNESP), Sao Paulo, Brazil.