Tracking the Consumption Decline

The new semester has begun, and I was reviewing economic trends in my macro courses. In my lectures, I highlighted the sharp drop-off in consumption. In the following, I discuss how well my predictions for consumption from last November have held up.
First, a recap of events:

Figure 1: Real consumption year-on-year growth (blue) and real GDP year-on-year growth (red), calculated as 4 quarter log differences. NBER defined recession dates shaded gray. Source: BEA, GDP 2009Q2 second release, NBER, and author's calculations.

Figure 1 shows how the year-on-year growth in consumption has hit very low rates, lower than at any period in the past forty years. So too has GDP growth (where now are all those people who were denying the depth of the recession, e.g. [0]?).

In Figure 2, I show individual component year-on-year growth rates. (Durables, nondurables and services are 10%, 21.8%, and 68.1% of nominal consumption, respectively.)

Figure 2: Real durables consumption (dark blue), nondurables (pink) and services (light green) year-on-year growth, calculated as 4 quarter log differences. NBER defined recession dates shaded gray. Source: BEA, GDP 2009Q2 second release, NBER, and author's calculations.

A log-levels depiction of aggregate consumption is of interest. In Figure 3 I show the log level of real consumption (blue line), converting consumption measured in Ch.2005$ into Ch.2000$ (I use the ratio of the base year 2000 and base year 2005 deflators in 2000 to convert one to the other).

Figure 3: Log real consumption in rescaled Ch.2000$, most recent vintage (blue), log real consumption in Ch.2000$ (red) and Nov. 13, 2008 forecast (purple). NBER defined recession dates shaded gray. Rescaled by using ratio of PCE deflator in 2005 to 2000. Source: BEA, GDP 2008Q3 preliminary release, GDP 2009Q2 second release, NBER, and author's calculations.

What is clear is the drop in consumption has been marked, especially when measured against the upward drift observed before the crisis. I've included consumption series reported in the October 30th GDP release, as well as my forecast from my November 13 post. In that analysis, I used error correction models assuming long run cointegration between consumption components (durables, nondurables, services), GDP and equity and nonequity wealth to forecast out-of-sample assuming certain paths for GDP and wealth.

It's hard to tell (because the benchmark revisions added to the level of consumption, as they did to GDP [1]), but the actual decline from 2008Q2 to 2009Q2 was 1.8% (log terms), vs. my prediction of 2.8%. This is a little surprising since GDP had decline 4% instead of the 1.6% assumed, and from 2008Q2 to 09Q1, real net worth had declined by 15.8% instead of 15.1% I assumed.

Figure 4: Log real household net worth in Ch.2000$, latest vintage (blue), log real household net worth in Ch.2000$ from November 2008 (red); red series observations after dashed line are November 2008 forecasts. Series deflated by PCE deflator. NBER defined recession dates shaded gray. Source: Federal Reserve, Flow of Funds, BEA, NBER, and author's calculations.

Of course, I used a particularly simple model of consumption, where the key determinants were equity and nonequity wealth and GDP. The difference in behavior between GDP and disposable income (keeping in mind the tax rebates in 2009Q2) might account for the higher than predicted consumption (expressed relative to 2008Q2). Using Romer’s estimates of $40 billion in rebates, an impact multiplier of 0.5, and (unlike [2] and [3]) remembering to use consumption at quarterly rates, one finds that this effect could account for 0.9 percentage points of the 1.0 percentage point mis-prediction (although if I went through the trouble of updating the forecast to account for the actual evolution of GDP and net wealth, I'd have a bigger gap to account for).