We are very sorry to be the bearer of bad news, but your pay for the next year or two has been cut significantly. What do you want to do about it? If you have enough savings to cope with it, then you can use those to bridge your consumption until better days. But if your bank account is not loaded or you are not sure how long your salary cut will last, then perhaps you would consider cutting spending. But where would you start? Well, your partner’s words of wisdom are: “we do not have to go out for dinner every week”, “the glorious summer holidays in Puglia can wait another year”, “our old car is still ok to drive us around”, “why don’t we cancel our streaming subscriptions and gym memberships?” It all sounds very sensible and so this is what you cut.
The issue, however, is that during a recession, there are many people like you (and us, by the way) who are workers that, in the face of economic adversities, may prefer (or have no other choice than) to cut spending. And, obviously, we all start by cutting non-essentials. As a result, revenues in hospitality, sports and gym industry, manufacturing sectors, entertainment, and so on all start to drop and, with them, labour demand in those industries falls too. But who works in those sectors? What do waiters, ushers, trainers, flight attendants, and blue collars have in common? They provide for their families, are typically on lower income, and use most of their salary to pay for bills and rents (or repay their mortgages). When the rich cut their non-essential spending, these workers see their earnings reduced because of weaker labour demand in the sectors producing non-essentials, where they work. The spending of the rich drives the income of the poor. And this hit to low-income workers leads, in turn, to a further reduction in both essential and non-essential spending. As a result of these second-round effects, the initial contraction worsens, and the recession deepens.
Empirical evidence
At this point, we wish we could add the disclaimer: “All characters appearing in this work are fictitious. Any resemblance to real persons, living or dead, is purely coincidental”. Unfortunately, we cannot. This ‘story’ is based on a paper (Andreolli et al. 2024) in which we advocate a new paradigm to think about consumption changes based on the distinction between essential and non-essential spending. We uncover three main empirical regularities in post-WWII US data: (1) during recessions, spending on non-essentials contracts significantly more than spending on essentials; (2) labour earnings in industries that produce non-necessities are significantly more cyclical than earnings in essential sectors; and (3) workers employed in non-essential industries tend to earn lower salaries than workers in essential sectors and are far more likely to be hand-to-mouth.
In Figure 1, we report the evolution of non-essential spending as a share of total expenditure over a post-WWII US sample. The trend reflects the well-known observation popularised by Engel (1857) that as a country becomes richer, non-essential consumption increases more than essentials. Less well-known, however, is the working of Engel’s law over the business cycle: whenever the US economy contracts (as exemplified by the NBER recession-shaded areas), non-essential spending goes down, cyclically, as a share of total household expenditure.
Figure 1 Non-essential consumption share over time
Note: Personal consumption expenditure shares of non-essentials, constructed from chained (2000$) spending series, and as a proportion of total classified expenditure. Underlying data are from the BEA PCE by Type of Product tables. Shaded areas in grey represent NBER recession dates
Another way of telling the same (expanded) ‘story’ is reported in Figure 2. It tracks the evolution of spending and earnings in essentials and non-essentials during an average US recession, whose start is normalised to zero. The panels show that falls in aggregate consumption and earnings are actually driven by declines in the non-essential sectors, which are far larger than the decline in their essential counterparts. These are findings (1) and (2) above, but this is not the whole story. In Figure 3, we show that the share of hand-to-mouth workers (i.e. households with low liquid wealth relative to income, following Kaplan et al. 2014) is far larger in non-essential industries, especially at the bottom of the income distribution (finding iii).
Figure 2 Sensitivity of essentials and non-essentials to recessions
Note: Response of essential and non-essential series after the start of a recession. To construct these responses, we first log and detrend the series using the HP filter (𝜆 = 14,440). For the earnings series, we report a six-month moving average to reduce noise. We then calculate the average decline in the series after all recessions, as defined by the NBER, between 1973 and 2007, which the data are available for.
Figure 3 Share of hand-to-mouth workers in essential versus non-essential sectors
Note: Panel Study of Income Dynamics (PSID). Sample: 2003-2021.
Monetary policy
The evidence above refers to a typical US recession, but ‘non-essential business cycles’ can be equally triggered by the actions of central banks. The first column of Figure 4 reports the effects of an identified monetary policy shock that moves the interest rate by 100 basis points on consumption and earnings, both in the aggregate and for each sector. While the aggregate effects resemble those in earlier macro studies, the sectoral ones are new: the decline in non-essential spending following a contractionary monetary policy is about two times as large and persistent as the decline in essential spending. But the sharpest heterogeneity emerges in the bottom row: the decline in non-essential earnings peaks significantly in excess of -4% whereas the drop in the earnings of the essential sectors is insignificant and never exceeds -1%. The last column reveals that the responses of essential and non-essential, for both consumption and earnings, are statistically different one from another.
Figure 4 Impulse response functions to a contractionary 100 basis point monetary policy shock: Consumption and earnings
Note: Blue lines are empirical impulse response functions (IRFs) to a 100bp increase in the 1y year government bond yields estimated by smooth local projections instrumental variable, where the instrument is the monetary policy shocks derived from the Gertler and Karadi (2015) high-frequency monetary policy surprises. Confidence intervals are reported at the 90% (dashed line) and 68% (solid line) level. Sample periods and controls for each column are specified in the main text and Appendix C.1. Red markers refer to quarterly IRFs from the estimated structural model of Section 4. Red “X”s correspond to variables that have been targeted in the structural estimation whereas red “O”s stand for variables that have not been targeted.
Model insights and policy implications
To interpret our empirical findings, we develop a two-sector New-Keynesian model with: (1) non-homothetic preferences, (2) hand-to-mouth agents, and (3) heterogeneity in sectoral labour composition. Households consume two types of goods that differ by their income elasticity: non-necessities are easier to postpone in the face of a negative shock. Workers have either low productivity and are hand-to-mouth or have high productivity and are unconstrained. Non-essential sectors employ a larger share of the former, as in the data. Because of non-homothetic preferences, workers with high productivity spend a larger budget share on luxuries. This implies that the intertemporal elasticity of substitution is heterogeneous across goods and households.
We estimate the structural model and perform counterfactual simulations that highlight two major results. First, the cyclical demand for non-essentials combined with a larger share of hand-to-mouth workers employed in non-necessity sectors makes the composition of labour demand endogenously cyclical: the labour earnings of the hand-to-mouth workers in non-essential industries decline significantly more during recessions. Second, the interaction between the cyclical spending composition (i.e. Engel’s law) and the heterogeneous sectoral labour composition accounts for a significant share of the effects of monetary policy on aggregate consumption and greatly amplifies business cycle fluctuations relative to models which do not incorporate these features. This amplification mechanism is novel and distinct from earlier studies (e.g. Acharya et al. 2023).
As for fiscal policy, our analysis begins by generalising the important theoretical result in Correia et al. (2013) that a suitable combination of consumption and labour income tax changes can exactly replicate the effects of monetary policy (see Bachmann et al. 2021, for an example of unconventional fiscal policy in Germany). We go further, by showing that, in our framework, a VAT change levied only on non-essentials (only on essentials) amplifies (dampens) the effects of fiscal policy relative to a VAT change that raises the same tax revenues but applies uniformly to all sectors. Interestingly, a hike (cut) only in the VAT rate of non-essentials hits (benefits) the earnings of low-income households more than a VAT change applied only to essentials or uniformly.
Conclusions
Households and workers differ greatly in their exposure to the business cycle along the income distribution. The composition of product and labour demand across essential and non-essential sectors is crucial to identify and quantify their cyclical exposure to shocks. In the face of economic adversities, non-essential purchases are easier to postpone, and their fall is driven by affluent households. The latter has a particularly large effect on low-income/hand-to-mouth workers, who are more likely to be employed in non-essential industries and therefore see their labour demand contract more: the higher cyclicality of non-essential spending leads to a higher cyclicality of non-essential earnings. This framework offers novel insights to understand business cycle fluctuations and the dynamic effects of stabilisation policies. For instance, we show that policy interventions targeted to non-essentials (such as sector-specific VAT changes) are likely to stimulate aggregate consumption (and benefit low-income households) far more than monetary policy and tax policies that either apply only to essentials or are levied uniformly on all sectors and households in the economy.
References
Acharya, S, E Challe and K Dogra (2023), “The Science of Monetary Policy Under Household Inequality”, VoxEU.org, 1 August.
Andreolli, M, N Rickard and P Surico (2024), “Non-essential business-cycles”, CEPR Discussion Paper 19973.
Bachmann, R, B Born, O Goldfayn-Frank, G Kocharkov, R Luetticke and M Weber (2021), “A Temporary VAT cut as unconventional fiscal policy”, VoxEU.org, 20 November.
Correia, I, E Farhi, J P Nicolini and P Teles (2013), “Unconventional Fiscal Policy at the Zero Bound”, American Economic Review 103: 1172–1211.
Engel, E (1857), “Die Productions- Und Consumtionsverhaltnisse Der Königsreichs Sachsen”, Zeitschrift des Statistischen Bureaus des Königlich-Sächsischen, Ministerium des Innern, [1895 repr. in Bulletin de l’institute international de statistique], v. 9, p. 1.
Kaplan, G, G L Violante and J Weidner (2014), “The Wealthy Hand-to-Mouth”, Brooking Papers on Economic Activity 45: 77-153.
