Tuesday, August 20, 2019

Wages and the Business Cycle 1


One of the trickiest things for newspaper economics journalists, it seems, is to try and sort out what should happen to wages during business cycle ups and downs.  Certainly, a lot of the American press at the moment seem to think that there’s something wrong with the present recovery because wages aren’t rising, or not rising as rapidly as the press had expected.  You do get recognition that wages are rising in certain sectors of the US economy, typically sectors like housing construction where the markets are local and fall into the category that Stock and Watson recently identified as cyclically sensitive (meaning, in essence, producing non-tradeables) but there seems to be a broad feeling in the US press (and to a degree the Canadian, although there seems to be growing awareness that what happens in the US isn’t automatically mirrored in Canada) that for a recovery as long as the present one (albeit a slow recovery), wages should have risen broadly.


It’s easy to criticize economics journalists, but we have to acknowledge that aggregate labour market analysis is not exactly one of the more settled areas of economic theory.  We’ve got micro labour market analysis pretty well sorted out, but aggregating across micro markets to get the macro market is not one of the more satisfying areas of economic analysis.

(Stephen Gordon has an interesting discussion of Canadian real wages at Worthwhile Canadian Initiative, showing, among other things, that it's a good idea sometimes to take a long horizon view of your data, and that it's always a good idea to keep the relevant theoretical concepts at the front of your mind.)

For a start, we have to draw a distinction between nominal and real wages.  And in doing that we have to acknowledge that nominal and real wages can move in opposite directions during business cycle swings.  We also have to draw a clear distinction between changes in wages over the business cycle and changes associated with economic growth.  And we have to explain about the value of the marginal product of labour.

The VMPL is at one level a fairly easy concept to explain.  If a firm is going to hire more labour (and here the additional labour can be in the form of more effort from a fixed number of workers or it can involve hiring more workers – that’s one of several empirical issues which bedevil the field) it won’t want to pay any more for that additional labour than bringing the labour on will add to its revenue.  So the highest wage it will pay is the VMPL – the addition to total revenue that the firm will experience as a result of having hired that additional labour.  Since this is a marginal product concept, we’re holding all other inputs constant – not just capital but also materials.  Empirically that might never happen, since materials might well move pretty much in lockstep with output, but conceptually it makes sense and we can adjust for fluctuations in other inputs when we go from the theoretical to the empirical.

(The best way to think about the VMPL might be in terms of something like a Tim Horton’s in the short run.  The capital stock is fixed, the materials – coffee etc. – is fixed, and its labour which adjusts through the day as the demand for the product adjusts.  At any given time of day, facing the demand associated with that time of day, the store has to decide how much labour it needs, which also involves deciding how much labour would be too much.  Adding labour increases the number of cups of coffee it can supply, but the productivity of additional labour is limited by the amount of capital and materials in the store.  There’s no point bringing in more labour if there are so many workers already there that the new ones can’t get anywhere near the equipment.)

We summarize the firm’s decision by saying that the nominal wage will be equal to the value of the marginal product of labour or, dividing through on both sides by the price of the product, the that real wage (in product terms) will equal the marginal physical product of labour (the MPPL).  The important thing to remember about the MPPL is that it’s an engineering concept, not an economic one.  It falls out of the nature of the production process – mathematically it’s the derivative of the production function with respect to the labour input.  The economic concepts are the nominal wage and the price.  The MPPL only shifts when something in the technology shifts – either we add more capital for the labour to work with or the technology of production changes.  So long as we’re holding the capital stock and the technology of production constant, the MPPL curve won’t shift.
Now think about sticky-nominal-wage explanations for unemployment.  Let the demand for output fall.  We usually assume that this means that the aggregate price level has fallen.  We also assume that there’s an aggregate counterpart to the industry level VMPL curve (although we recognize that in trying to estimate it we have to allow for the composition of the aggregate economy).  The sticky-nominal-wage models assume that, if the nominal wage were to fall along with the price level, the real wage would remain unchanged, meaning that the profit maximizing level of employment would remain unchanged, and there wouldn’t be any real effects from the downturn, only nominal ones.  If the nominal wage doesn’t change the real wage will rise as the price level falls and the aggregate labour market will slide along the aggregate MPPL curve.  That aggregate MPPL curve is, like the industry-level one, an engineering concept, and by going over to real wage terms we’ve given it no reason to shift.

But if the real wage rises and the economy stays on the aggregate MPPL curve, the level of employment must fall, meaning that the level of unemployment must rise.  So when we go into a recession, the real wage must rise.  And when we come out of a recession, the sticky-nominal-wage theories assume that the price level will start to rise, bringing the real wage down and sliding the economy along the aggregate MPPL curve in the direction of higher employment (and hence lower unemployment).

So the s-n-w models predict that the real wage will be counter-cyclical, rising as we go into a recession and falling as we go into a recovery.

These models actually allow the nominal wage to change, so long as it lags behind the price level.  Thus when we go into recession wages can fall, but for unemployment to rise wages have to fall proportionally less than the aggregate price level falls.  And as we come into a recovery, nominal wages can rise but for employment to rise the increase in nominal wages has to be proportionally less than the rise in the price level.  In other words, real wages have to fall.

That’s one of the most controversial predictions of the s-n-w models: that real wages should be falling as we move into a cyclical upswing.  You can see why economic journalists would have trouble with it. There’s a natural, gut feeling that as the economy comes out of recession, more real, physical stuff is being produced, so the amount of real, physical stuff which goes to each unit of labour should also be increasing.

And it’s one of the trickiest ones empirically.  The empirical literature on the cyclical nature of real wages is one which itself just seems to cycle around, never reaching a conclusion.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.