Auto sales have recovered to the 16.5-17 million range, and many observers predict further gains in coming years (despite, as we previously noted, missing expectations for the last few months) [9]. But to Goldman Sachs, the current sales pace already looks high relative to the medium-term fundamentals; and their assessment of scrappage rates, population growth, licensed drivers, and vehicle ownership suggests that trend demand for autos - excluding cyclical fluctuations - is only 14-15 million units per year.
Goldman Sachs' Jan Hatzius explains...
Auto sales have recovered nicely from the crisis and now stand in the 16.5-17 million range. Many observers expect further gains in coming years. For example, the consultancy IHS Automotive Insight--a respected source of industry analysis--believes that sales will stabilize at just over 17 million through the early 2020s.
Are these predictions realistic? To answer this question, it is helpful to break down the level of auto sales into two components, the number of autos needed to replace scrappage and the growth in the stock of autos on the road.
(1) auto sales = auto stock x (% scrappage + %ch in auto stock)
We start with the scrappage rate, shown in Exhibit 1. It is quite noisy in the short term but has fluctuated around a slowly declining trend over the long term. At the moment, that trend stands at 5% or a bit less, which implies replacement demand of 12.5-13 million per year given the current auto stock of just under 260 million.
[10]
Exhibit 1: Scrappage Rate Has Gradually Trended Down and Is Now Just Under 5%
Next, we look at the change in the stock of autos on the road, which we can write as:
(2) %ch in auto stock = %ch in adult population + %ch in licensed drivers/adult population + %ch in autos/licensed drivers
The Census Bureau estimates that the adult population will grow 1% per year through the end of the decade. But this probably overstates the growth in the auto stock because the other two terms in the equation are likely to be negative.
First, the left-hand panel of Exhibit 2 shows that the share of licensed drivers in the adult population has fallen by almost 0.5 percentage point per year since 2008. The fact that the decline started in 2008 raises the question of whether it could be due to the Great Recession. This may be a factor, but the right-hand panel of Exhibit 2 suggests a more structural explanation. There has been a long-standing trend decline in licensed driver shares among younger age groups. Until a few years ago, this decline was offset by an increasing licensed share among older individuals. But that increase has now ended so that the decline among younger groups is dominating the aggregate trends. Even assuming that some of the recent weakness is cyclical, we therefore expect the licensed share to decline by 0.2-0.3pp per year.
[11]
Exhibit 2: Licensed Drivers Make Up a Shrinking Population Share
Second, the trend in the number of vehicles on the road per licensed driver has changed recently. As shown in Exhibit 3, this ratio rose sharply until the early 2000s but has declined since then. Although it is uncertain how much of the weakness is due to cyclical versus structural forces, we would pencil in a mild downward trend of perhaps 0.1 percentage point per year in coming years.
[12]
Exhibit 3: Ratio of Vehicles to Licensed Drivers Seems to Have Topped Out
These estimates imply that the stock of autos on the road is likely to increase at a trend rate of about 0.6-0.7%, or 1.5-2 million per year. Combined with our estimates for replacement demand, this implies an underlying trend rate of auto sales in the 14-15 million range, or about 2-2.5 million below the recent pace. (Longer term, the risks to these estimates are probably on the downside, especially if the "sharing economy"--exemplified by companies such as Zipcar and Uber--makes deeper inroads into the transportation sector. But that is a different story that is not central to the analysis presented here.)
How big an issue is all this for the US economic recovery? Although the GDP implications are negative, they are quite small. Consumer spending on motor vehicles and parts accounts for 2.5% of GDP, but only about half of this consists of purchases of domestically produced new light vehicles (the rest is imports, used vehicles, and parts). Moreover, because of its medium-term nature, it is natural to spread the decline implied by our analysis over a fairly lengthy period. If it occurs over the next 4-5 years, the drag on annualized GDP growth would average only about 0.05 percentage point per year.
* * *
Hatzius is careful not to entirely destroy hope though, as he concludes,
But there is also a more structural difference between housing and autos, namely that long-term population growth is likely to be much more tightly linked with housing demand than with auto purchases. After all, everyone needs a place to live, but more people are finding that they do not need a car.
