The Hancock curve

In the acracy post I briefly touched on the apparently influential idea that there’s an economic calculus you can apply to COVID-19 – an inverse relationship between R and GDP growth. I didn’t really want to go into it much because I was writing about acracy rather than economics, so let’s get back to it.

When government “sources” talked about “running it hot” for the sake of the economy, they were asserting an implicit model where there is a correlation between the spread of the virus and the state of the economy. Quarantine measures, this asserts, harm economic activity, and therefore pushing down the value of Rt comes at a cost in terms of growth. It therefore remains to estimate the parameters from the data and pick your poison.

If this is familiar, it’s because it’s state of the art macroeconomics from about 1965. The famous Phillips curve describes an empirical inverse relationship between wage inflation and unemployment (itself a proxy for growth) originally based on data from 1913 to 1948. This is William Phillips’ original plot:

Over time it developed from being a data insight into being a policymaking and forecasting tool, shifting its meaning from just an observation that low inflation and high unemployment were correlated to an assertion that if you wanted to target an x% unemployment rate you needed to tolerate y% inflation.

There were and are good reasons for the relationship to exist. Unemployment weakens workers’ ability to bargain over wages; at full employment, increasing production by pulling in more workers is difficult, so consumers tend to bid up prices. The problem, though, is that the Phillips curve relationship turned out not to be stable – the parameters of the curve could change. A huge amount of economics since the 1960s went in to arguing about whether the curve still existed, what were good estimates of its current location, and why it was shifting. I’m not going to go into all this now, but let’s just keep the point that the curve relationship cannot be assumed to be stable.

The Hancock curve, to name it for Matt Hancock, jumps right past the original phase of empirical observation and cuts straight to policymaking, because there’s no counterfactual world without quarantines, supply-chain disruptions, or closed borders in which we can observe the relationship of an unconstrained outbreak with GDP. (And a good thing too.) It’s not really a model of the interaction between the virus and the economy, but rather of the economic impact of quarantine restrictions. The whole idea is that we can look at the level of restrictions generating a given Rt, read off the economic impact, and decide whether it’s worth it.

The problem, though, is that there is no reason to think the relationship is stable. Unlike a real Phillips curve, one advantage this one has is that there is a well-defined equivalent of NAIRU, the supposed non-accelerating inflation rate of unemployment Milton Friedman spent the 80s searching for. It’s Rt=1, the point where the daily case numbers are constant. The problem, though, is that the case where Rt=1 and daily cases = 10 is very different from Rt=1 and daily cases = 26,000.

In the first regime we would be justified in thinking the outbreak was well under control, and probably reasonably confident about the economic future. This is, in fact, what we observe in countries that have got a grip on the virus. The Hancock curve would be set well to the right on the chart. In the second regime, though, we would be justified in thinking that the virus is going to get us in the end. We would therefore be unwilling to take any economic risks. The Hancock curve would be set well to the left on the chart. Further, if Rt strayed over 1 and case numbers rose, the curve would shift leftwards. Any slip can send the system zooming out of control. (Keen and agile minds may notice this is now an expectations-augmented Hancock curve.)

It’s been pointed out by cleverer people than me that in the end, although the monetarists of the 70s and 80s denied the existence of a Phillips curve relationship in theory, their policy advice in practice amounted to choosing mass unemployment in order to crush inflation, a Phillips curve trade-off with a vengeance, and in fact that was how generations of politicians communicated what they were up to. Both the economics and the politicians also added a further twist, though, which was the promise that the “short-term pain” would lead to future economic growth.

You’d think this would be something a bunch of Tories would remember.

Update: this isn’t really a Hancock curve because it’s a snapshot international comparison rather than a time series but it does support the conclusion that the best idea would be to go full Paul Volcker on the virus.

4 Comments on "The Hancock curve"

  1. Excellent post.
    I liked a lot that you’ve made the point that the scale of current infection matters re: R and cases.


    1. this is something I said to Tony Yates on the twitter months ago – you can only catch $VIRUS from another sufferer, so if n=0, R is entirely academic and Rt must be zero as well. on the other hand, the risk you face as an individual is very much linked to scale. Rt=1 with tens of thousands of cases a day is literally “the virus is spreading steadily through the population and will get you in the end”.


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.