Sumner confesses that he has begun looking at unemployment as a leading indicator, rather than as a lag as is usual. He then goes on to show why this is a useful way to look at unemployment, but doesn't discuss how it could be that something economists and the business press think of as a lag is actually a lead. I'm going to take a stab at it, and do so without arguing we are in a Keynesian world where recovery doesn't happen until people have money to spend.
Thinking of unemployment as a lag is natural. Assuming there are costs in hiring and firing, there is likely a threshold level of change to economic conditions that has to be met before a firm reacts and either hires or fires. Sales go down, inventories pile up, and only then does the firm start letting people go.
From a rational expectations framework, however, we have to assume firms are forward-looking. In the definition of a lagging indicator here, they even claim that unemployment lags output because firms are forward looking. But that means they don't wait for inventories to pile up. When bad news comes that will likely affect sales, it makes sense to cut back a little on hours now so that the inventories don't even pile up. This is an increasingly important problem as inventory management has become a key factor for many firms' operations. Thus it wouldn't be actual changes that matter, but expected changes, and we're back to Sumner's world where expected inflation and expected Fed actions are what matter. In other words, unemployment ought to LEAD because hiring and firing decisions will be made in advance of conditions.
One difficulty is that unemployment is an aggregate of three decisions (hiring and firing by firms and entering and exiting the labor force). If fires remain constant while hires slow down, unemployment increases in advance of bad news.
My guess would be that hires are a better leading indicator than fires, and fires a better leading indicator than entry and exit from the labor force. People start looking for jobs once they receive positive signals that they could get hired, but if they haven't been looking it may take some time for them to receive those signals. Put together, unemployment The Aggregate behaves like a leading indicator as firms adjust to expectations but will behave like a lagging variable as discouraged workers come back.
Anyone in favor of breaking unemployment down into these three pieces and seeing how each flow corresponds with various measures of NGDP?
[IE - I am assuming that firms are more forward looking because they are constantly gathering more information about future conditions while the discouraged unemployed likely face some information accumulation costs. I am also assuming that the monetary and non-monetary costs of hiring and not hiring are lower than for firing and not firing. It is cheaper for a firm to put off hiring someone if the news might turn bad in the future than to pro-actively fire someone.]