John Kemp's Profile
Doing the inventory swing – Sources of GDP growth in the year after recession
With a lot of fog drifting across the analytical battlefied, we need to start distinguishing between two statements:
(1) The economy is approaching the cyclical trough;
(2) A strong sustained recovery is set to get underway.
There is good evidence to support proposition (1), but it does not necessarily imply proposition (2). It is possible to be optimistic the worst of the downturn may now be over (or very nearly so) while remaining cautious about the prospects for recovery when the cyclical turning point is passed.
To understand the point, think back to the last recession. The National Bureau of Economic Research (NBER), official arbiter of the U.S. business cycle, dates the cyclical trough to November 2001 (eight months after the peak in March 2001 — and just two months after the attack on the World Trade Centre in New York). But it was not until more than two years later that there were signs of a strong and sustained recovery.
The fitful nature of the expansion during 2002 and 2003 is why the Fed kept rates low for so long. In fact, the Fed was still cutting interest rates to 1% in June 2003 because the expansion was so anaemic, and did not feel confident to begin raising them from this ultra-low level until June 2004. Uncertainty caused by impending war between the United States and Iraq was one factor holding the expansion back, but not the only one. The recovery failed to become “self-sustaining” for almost three years.
The same could easily happen again.
The attached chartbook is an attempt to look beyond the downswing of the cycle at how the U.S. economy behaves in the first twelve months of recovery.
First a quick technical digression. As I have explained before, the national income and product accounts (NIPAs) of which gross domestic product (GDP) is the focal point add up domestic production in three ways: by summing expenditures, by summing incomes, and by summing production.
In analytical terms, the three approaches are identical (for every unit of output there are corresponding incomes and expenditures). But as a practical matter, it is far easier and faster to estimate GDP by adding up expenditures and working back to income and output. The federal government makes separate estimates based on incomes and output when tax returns and the quinquennial business surveys become available later (and then has to adjust all the measures to try to make them consistent).
The only problem with this expenditure –> output approach is the need to adjust the data to take account of items produced in one period but consumed in another (ie changes in the amount of produced but unsold raw materials, work in progress and finished products held by manufacturers, distributors and retailers as “inventories”).
Inventories are one of the most volatile components of the output-income-expenditure system and a key driver of cyclical behaviour.
For analytical purposes we can separate GDP growth into two components:
Growth in GDP = Growth in final demand (consumer spending + business investment + government spending + exports – imports) + Change in the Level of Inventories.
Large inventory changes are usually unsustainable and tend to be quickly reversed. A large build up in inventories one quarter is normally followed by an equally sharp decline in the following one (see Chart 3 in the chartbook).
Charts 1-3 and Table 1 look at how these three measures (headline GDP, final demand, and inventories) have behaved on a quarter-to-quarter basis since 1948. For clarity, I have reported all growth rates as simple quarter-to-quarter rates, rather than the quarter-to-quarter annualised ones that are normally used by the Bureau of Economic Analysis, since this is more useful for the type of quarterly analysis we are looking at here. I have reverse engineered the annualised data by the simple expedient of dividing by four — which is not strictly accurate but good enough for our purposes here, especially since the percentage changes are small.
The immediate point is that headline GDP and final demand are much more stable than the inventory component. Headline GDP and final sales have been negative in only 37 and 35 quarters respectively out of a total of 245 since 1948. In contrast, inventory changes subtracted from GDP about half the time (119) and added to it roughly as often (126).
Of the two main components, final demand was slightly more stable than GDP (reflecting the volatility of inventories). The average increase/decrease in headline GDP between one quarter and the next (ignoring signs) was 1.06 percent, a slightly larger change than the average change in final demand (0.95 percent). Headline GDP changes have a slightly higher standard deviation (1.03 percentage points) than final demand (0.85 percentage points) confirming that headline GDP is more variable.
Before anyone points out inventories (with a mean change of 0.48 percent and standard deviation of 0.67) are more stable than either — this appears true, but recall that inventories are only a very small component of GDP and this is the contribution which they make to the total growth/fall each quarter — so they are actually a source of enormous volatility for their size.
Armed with this understanding, we can start to analyse the recessions since 1948.
I am going to divide them into two groups.
The first group is “severe recessions” in which FINAL DEMAND fell for at least two consecutive quarters. There are four of these ending in 1954, 1974, 1982 and 1991 respectively.
The second group is “other recessions” where final demand remained positive (most of the time) but the economy was pushed into recession (two or more quarters of negative growth) by changes in the inventory component. There are five of these “other recessions” or “inventory recessions” ending in 1949, 1958, 1960, 1970 and 2001 (1960 and 2001 do not strictly fit the pattern since there were at least two quarters of negative growth but not consecutive, but they are near enough to be included and both are generally regarded as being recessions).
So now we have our “deep recessions” and “inventory/other recessions”, we can look at how the economy behaved in the four quarters after the cyclical trough was reached in terms of both growth in final demand and inventory rebuilding.
There are crucial differences.
In the case of deep recessions, recovery was led by final demand. Inventory rebuilding played a role, but only after a lag. After all four deep recessions, inventory changes actually subtracted from GDP growth in the first quarter after the recession was officially over. They made a small positive contribution in the second quarter of the recovery in three of four cases, but only a modest one. It was not until the third quarter after the recession ended that they gave a significant boost to GDP.
The pattern after inventory/other recessions is very different. In four of the five cases, inventories made a positive contribution to GDP in the first quarter after the recession ended (the exception was the anaemic recovery after the 2001 recesssion, and the lack of a stronger inventory response here may have contributed to the failure to achieve self-sustaining recovery in this case). The inventory-driven recovery fell away somewhat in the second quarter before becoming more resurgent in the fourth. In contrast to the deep recessions, growth in final demand was much less important in the early months of recovery, with inventory rebuilding supplying most of the initial impetus for expansion.
To summarise, recoveries from deep recessions are led by final demand, with inventory rebuilding not providing a significant source of stimulus for six months. In contrast, recoveries from inventory/other recessions are led by inventory rebuilding, with final demand playing a more limited role until later, though the initial boost from inventories can dip in the second and sometimes third quarter after recovery begins.
So what type of recession and recovery are we in this time?
A quick look at Chart 2 reveals that the current recession is a deep recession that has seen significant destruction of final demand. So it is part of group 1.
If this is the correct characterisation, inventory changes might play only a limited role in the early stages of the recovery. Assuming the economy hits the cyclical trough at the end of Q2 or sometime in Q3, inventory changes might not play a significant stimulative role until Q1 2010.
One caveat: our sample size of post-1948 recessions is very small, so we should be careful about drawing strong conclusions.
But the experience of inventory-driven and final-demand driven recessions over the last sixty years suggests we should be cautious about pencilling in a strong inventory-driven rebound in H2 2009.