One of the startling things about consumption behavior is that, despite the burst of spending surrounding the holiday season, per capita consumption in 2011Q3 has only re-attained the levels of 2008Q3. Various explanations have been forwarded, ranging from the failure of Keynesian economics , to the decline in income prospects or higher income uncertainty, or to the decline in observed net worth  (deleveraging, in certain interpretations).
Figure 1 shows log GDP and consumption rescaled to 2009Q2=0.
Figure 1: Log GDP (blue) and consumption (dark red), Ch.2005$, SAAR, rescaled to 2009Q2=0. NBER defined recession dates shaded gray. Source: BEA, 2011Q3 2nd release, NBER, and author’s calculations.
Taylor has argued that the failure of consumption to rise with income tax cuts is proof of the failure of the simple Keynesian consumption function, viz.
C = a + bYd
Where C is real consumption, and Yd is disposable income.
Of course, nobody really takes the simple Keynesian consumption function as a realistic depiction [Update: In his 2011 JEL article, Taylor incorporates a wealth effect.]. As far as I know, all large scale macroeconometric models incorporate a wealth variable.
C = a + bYd + cW
Where W is real wealth.
In this perspective, the decline in consumption, despite the maintenance of disposable income, is no surprise, given the massive decrease in net worth starting in 2007.
Figure 2: Log consumption, Ch.2005$, SAAR (blue, left scale), log household net worth in billions, deflated using PCE deflator (dark red, right scale). NBER defined recession dates shaded gray. Source: BEA, 2011Q3 2nd release, Fed via FRED, NBER, and author’s calculations.
There is some debate whether there is a distinction between the wealth effect and deleveraging. I think there is – the decline in wealth induces lower consumption, but higher income flow uncertainty and tighter liquidity constraints also affect consumption by alternative channels.
One point of interest is to think about the implications of Hall’s permanent income hypothesis. It’s important to recall that the model implies consumption service flows follow a random walk, rather than consumption expenditures. To the extent that service consumption expenditures most closely conform to the flow of services associated with consumption. Here’s a plot of the cumulative change in expenditure categories since 2009Q2.
Figure 3: Change in consumption by category, in billions of Ch.2005$, SAAR, since 2009Q2. Source: BEA, 2011Q3 2nd release, Fed via FRED, and author’s calculations.
The fact that services consumption has rebounded sharply in the last quarter (at 2% annualized) is consistent with a jump in perceived permanent income (or a big decrease in liquidity constraints).
There is a limit to what can be gleaned from time series. That’s why Atif Mian, Kamalesh Rao, and Amir Sufi have appealed to cross section data to identify the impact of deleveraging on consumption. From “Household Balance Sheets, Consumption, and the Economic Slump”:
Until recently, macroeconomic models often abstracted away from variation across households in balance sheet strength. In this paper, we show that weak household balance sheets played a central role in the dramatic collapse in aggregate demand from 2007 to 2009, and the continued weakness in consumption through 2011. We show that the presence of high levels of leverage implies that a large negative shock, like the shock to the housing market, has serious distributional implications. In particular, households with high debt balances experience the sharpest reduction in net worth when a large asset class such as housing loses value.
We also show that net worth declines resulting from high household debt and a collapse in asset values have serious implications for the real economy through their impact on consumption. Highly levered households cut back drastically on their consumption in order to repair their impaired balance sheets. Despite a sharp reduction in interest rates, there is no compensating increase in consumption by the non-levered households. The net result is therefore a large reduction in aggregate demand.
We quantify this effect in this paper through the use of disaggregated, county-level consumption data. Measuring consumption has long been a challenge, especially at a disaggregated level and over long time periods. We overcome these challenges with a novel data set on county and zip code level consumption of new automobiles, and county level consumption of durable, non-durable and grocery purchases. Using prior literature on the relationship between housing supply elasticity and house prices, we instrument the household debt levels with housing supply elasticity of the county.
Our estimates show that the household balance sheet channel is responsible for a very large fraction of the decline in consumption during and after the recession. Further, household balance sheet problems continue to depress consumption through 2011. The distributional implications of large wealth shocks in the presence of leverage cannot be ignored.
The core of the Mian-Rao-Sufi thesis is illustrated by the cross section patter of heightened sensitivity of durables spending to high leverage and big shocks to asset values (primarily in housing). For automobiles, see Figure 3. For a comparison of durables to essential goods (groceries), see Figure 4.
Figure 3 from Mian, Rao and Sufi (2011).
Figure 4 from Mian, Rao and Sufi (2011).
What are the policy implications? Although Mian et al. note that 65% of the job losses in the depth of the recession were due to balance sheet effects, they don’t make any specific policy recommendations. In my view, Torsten Slok’s assessment (“US Consumer Deleveraging: More adjustment needed,” December 2011, not online) provides a road map:
Most household deleveraging indicators are currently at their 2007/2008 levels. If they have to fall to their pre-housing bubble levels (=2003/2004 levels) then we still have several years of household deleveraging ahead of us. But how much more deleveraging to expect will depend importantly on what happens to equity and home prices. In other words, private consumption is not only driven by the liability side of households’ balance sheet but also — and perhaps most importantly — by the asset side.
In other words, working to reduce the indebtedness of highly leveraged households, either through formal restructuring or more rapid inflation, and supporting asset prices, will be necessary. Monetary policy can help support asset prices, but so too will faster economic growth which would accompany greater fiscal stimulus.
This post originally appeared at Econbrowser and is posted with permission.
2 Responses to “Consumption: Distinguishing Between Keynesian and Permanent Income Motivations, and Deleveraging”
One might also note that the notion of "maintenance of disposable income" is rather crude. If bankers have made more money, while 10% of the labor force is thrown out of work, how would that affect aggregate consumption? I think consumption would go down. So I would recommend an additional analysis where disposable income -maintenance or nonmaintenance- is broken down by propensity to spend / consume among different classes of wage earner.
I would like to thnkx with the efforts you’ve put in writing this web site. I am hoping precisely the same high-grade article from you finding out sometime soon too. In truth your resourceful way with words-at all offers inspired everyone to have my own site planning now. Genuinely blogging will be scattering its wings as well as developing rapid. Your current jot down is a good example.