Transmission of Monetary Policy with Heterogeneity in Household Portfolios

Households choose a portfolio between a liquid asset (bonds, tradable without friction) and an illiquid asset (real capital, tradable only with a certain probability each period but paying a higher equilibrium return), which “enable[s] the model to endogenously generate the distribution of portfolio shares and marginal propensities to consume across households as documented for the United States” (Introduction, p. 2). Aggregate investment follows q-theory, which the author contrasts explicitly with Kaplan, Moll and Violante (2018): “KMV do not have a Fisher effect on bonds and assume that part of dividends and profits are paid into the illiquid account. In my model, in contrast, aggregate investment follows q-theory” (Introduction, p. 3-4).

“The direct effect of the policy rate explains only one third of the total change in consumption, while equilibrium changes in income account for the remaining two thirds. The policy rate, however, still explains 86% of the investment response” (Introduction, p. 1). Aggregate investment falls by one-third less, and output falls somewhat less as well, in the economy with portfolio heterogeneity than in an otherwise identical economy with a representative portfolio, while consumption falls by more “because a sizable fraction of households are wealthy-hand-to-mouth and have high marginal propensities to consume” (Section 3.1, p. 17).

Because random participation in the illiquid asset market breaks the arbitrage condition that would otherwise force liquid and illiquid returns to move one-for-one, the “endogenous premium of the illiquid asset over and above the return on the liquid asset falls in response to a monetary tightening,” producing “incomplete pass-through of the policy rate to the return on illiquid capital, which stabilizes investment” (Introduction, p. 2; Section 3.1, eq. 19, p. 17). The mechanism is redistributive: “a monetary tightening increases income and wealth inequality… the liquidity premium falls in equilibrium until wealthy households, who have a low marginal value of liquidity, are willing to hold a larger fraction of outstanding liquid assets” – in the baseline calibration this is a 16-basis-point (annualized) fall in the liquidity premium (Section 3.1, p. 17).

Following Auclert (2019) and Kaplan, Moll and Violante (2018), the paper totally differentiates aggregate consumption and capital demand functions with respect to the equilibrium price sequence {liquid return, illiquid return, wage, profits}. For consumption, “the decline in wages explains around 40% of the initial fall,” with a persistent negative effect over 16 quarters, lower illiquid returns contributing similarly, while “the change in the liquid return… explains only 30% of the initial decline in consumption, and this effect is very short-lived” (Section 3.2, p. 18) – compared to Kaplan, Moll and Violante’s (2018) finding that the liquid return explains 20% in their model. For investment, “the direct effect of the liquid return via arbitrage explains 86% of the aggregate investment response,” with low wages contributing to the fall and high profits partly offsetting it (Section 3.2, pp. 18-19).

“Households close to the borrowing constraint have the highest marginal propensity to consume, around 40%, but the fraction of an additional dollar that goes into investment is less then 5%. Households above median wealth, in contrast, invest twice as much of any additional dollar” (Section 3.2, p. 20, Figure 4). Because wealthy households have a comparatively high MPI and poor households a comparatively high MPC, “as wealthy households have a higher marginal propensity to invest, they stabilize investment demand as they get richer through redistribution” – linking the wealth gradient in MPI directly to the paper’s aggregate investment-stabilization result.

Table 3 shows that labor income for households below median wealth falls by more than 1% of their consumption, while the top wealth quintile enjoys rising profit income (a disproportionate share are entrepreneurs) even as labor income falls economy-wide; “overall, a tightening of monetary conditions increases inequality because it redistributes from borrowers to savers and from households that earn wage income to those that earn profit income” (Section 3.2, pp. 19-20). Figure 5 shows that for low-wealth households, who rely on wage income and have high MPCs, “the fall in wages explains 80% of the consumption response,” and these households “even reduce their liquid wealth and portfolio liquidity… to smooth consumption,” while wealthy households increase liquid holdings – the differential portfolio response documented empirically in Section 4.

With sticky prices, surprise deflation from a monetary tightening redistributes from (high-MPC, low-MPI) borrowers to (low-MPC, high-MPI) savers, and “the Fisher channel amplifies the aggregate effects of monetary policy through aggregate demand”; quantifying this against a counterfactual real-debt economy, the author finds “aggregate output falls by roughly 9% more in the baseline model with nominal debt” (Section 3.2, p. 22). Under flexible prices, however, “heterogeneity in propensities to consume does not matter because any shortfall in spending is offset by firms lowering prices,” so “only heterogeneity in the marginal propensity to invest matters… a monetary tightening leads to an investment boom with flexible prices as inflation redistributes from borrowers with low to savers with high marginal propensities to invest” (Section 3.2, p. 22) – a sign reversal that isolates exactly which heterogeneity (MPC vs. MPI) is doing the work in each pricing regime.

Recalibrating the model to a U.S. post-2008 environment by setting the steady-state return on liquid assets to zero raises the liquidity premium from 2.5 to 4.5 percentage points and increases the wealth Gini by 2.6%, because “wealth-poor households, who predominantly save in the liquid asset, are less well-insured, while wealth-rich households… get roughly the same return as before.” In this higher-inequality steady state, “higher inequality amplifies the importance of the redistribution channel and, hence, further increase[s] the consumption response while decreasing the response of investment” (Section 3.2, p. 23) – so the paper’s core mechanism is itself state-dependent on the prevailing level of wealth inequality.

Regressing repeated cross-sections of household portfolios from the Survey of Consumer Finances on identified monetary policy shocks, the author finds that “when rates increase, households below median wealth reduce their portfolio liquidity, while households above median wealth increase portfolio liquidity” (Introduction, p. 3; Section 4). The model can only replicate this differential sign pattern when capital is imperfectly liquid: “as I increase the liquidity of capital, fewer households lower their liquid savings in response to a monetary tightening. When capital becomes perfectly liquid, all households increase their portfolio liquidity, which is counterfactual, since more than 50% of households lower portfolio liquidity in the data” (Introduction, p. 3).

Using two measures of the liquidity premium – Gomme et al.’s (2011) return on capital and the return on housing, both net of the risk-free rate – and local projections on monetary shocks identified via the Romer and Romer (1989) narrative approach, the author finds “both premia fall in response to a monetary tightening in line with the predictions of the model” (Introduction, p. 3; Section 4). This is offered as direct time-series corroboration of the theoretical mechanism (falling liquidity premium after a tightening) that drives the paper’s investment-stabilization result.

The author frames the paper as complementary to KMV: “they also decompose the effects of monetary policy into direct and equilibrium effects but focus on heterogeneity in MPCs. The key difference between both models is the portfolio problem: KMV do not have a Fisher effect on bonds and assume that part of dividends and profits are paid into the illiquid account. In my model, in contrast, aggregate investment follows q-theory” (Introduction, pp. 3-4). The paper’s stated distinct contributions are “new evidence for the importance of modeling assets with different degrees of liquidity” and being “the first to discuss the importance of heterogeneity in propensities to invest for monetary policy” (Introduction, p. 4).