In fact, I agree with DeLong and don’t even have a clear idea who Say was (don’t know his first name do know he wrote before Ricardo). In standard economist speak
Say’s law states that output can’t be below the quantity that people want to supply given wages and prices, that aggregate demand must rise to equal aggregate supply.
This is frequently assumed in economic theory. For decades there has been a heated debate among macroeconomists as to whether it is approximately true of actual economies.
DeLong argues (for example here) that we know that Say’s law must be false because the demand increase due to dot.com mania and the housing bubble caused temporarily increased GNP. The argument, basically, is that the governments money is as good as anyone elses and if government spending can’t drive up GNP neither can any spending.
I critique DeLong’s argument after the Jump.
Say’s law does not imply that government spending can’t drive up GNP (I’ve noted this repeatedly and it is a well known result in the literature). Say’s law also doesn’t imply that all spending must have an equally low correlation with GNP.
First, government spending can make us richer, even if Say’s law holds, if public capital is more productive than private capital. I will not discuss this possibility. Second government spending can make us poorer. It will make us poorer if it is wasted or if government investment produces less valuable capital than private investment. An increase in scheduled government spending is, thus bad news for people. If leisure is a normal good, people respond to this bad news by offering more hours of labor. If Say’s law holds, someone must take them up on the offer so hours worked and production increase.
The many economists who predict 100% crowding out are not just asserting Say’s law. They are also asserting that labor supply is inelastic. It is very odd that they assert this. However, assuming that labor supply is elastic matters little for the policy debate. Under standard assumptions increased labor supply can cushion the blow due to increased public spending, but it is still a blow.
DeLong’s interesting claim is that the same must be true of the dot.com bubble and the housing bubble. This is just not true.
First it is possible to have GNP fluctuations even if Say’s law holds. This is the original real business cycle (RBC) argument based on the idea that output is a function of labor, capital and technology and technological progress is stochastic and irregular.
First assume workers have an instantaneous utility function which is the sum of a function of consumption and a function of leisure. This means that exogenously increased wealth (like winning the lottery) causes increased consumption and increased leisure (reduced labor supply) so leisure and consumption are both normal goods. The rate of growth of consumption is a function of the real interest rate and the level of consumption (except in special cases where it depends only on the real interest rate).
Now assume that investment automatically equals saving minus the public deficit (Say’s law) and consider a growth model in which GNP is produced by a production function. Assume technological progress is labor augmenting so output is a function of capital and effective labor (technology times labor).
First consider steady technological progress. In this case the ratio of capital to effective labor converges to a constant (call it k*) and the economy converges to a balanced growth path. OK what happens if the economy is on the balanced growth path and there is surprisingly rapid technological progress ? The ratio of capital to effective labor is lower than k* so the interest rate is higher. The salary per one unit of effective labor is lower than it was on the balanced growth path, but the salary per unit of plain labor is higher (it increases less than proportionally to the increase in labor augmenting technology).
The increased real wage causes increased consumption for given labor supply. The increased real interest rate causes the rate of growth of consumption to increase. For some utility functions, the high real interest rate causes capital accumulates quickly so the capital effective labor ratio is restored. For one utility function (the pleasure due to consumption is logarithmic in consumption) per capita labor supply returns to the original level when the capital/(technology times population) ratio returns to trend. This is not a new model.
Brad’s point is that there was an additional effect on GNP due to the bubbles. First overestimates of the amount of technological progress due to the internet caused an additional increase in employment and output beyond that due to the true technological progress. One could interpret the housing bubble as overestimats of the value of the services which houses would provide in the future (not the present as rents were observed) and therefore overestimates of the efficiency of housing services production via construction.
Brad’s claim must be that true technological progress can cause increased employment, but that the false illusion of technological progress can’t. Huh ? Why not ?
What matters for the model is what firms are paying in wages to workers and interest rates to savers. For labor supply and output it doesn’t matter if the firms are living in a dream world and the bubble will burst.
Now the RBC models assume rational expectations, so the bursting of the bubbles would convince me that such models are terrible approximations to reality (except I always believed that) but the expansion in response to a perceived improvement in technology is totally exactly what the models predict.
I guess DeLong’s point is that the path of employment can’t be explained as the result of optimization of a function of consumption and leisure given actual data on hours worked, consumption, real wages and real interest rates (now observable with TIPS). I agree, but that has nothing in particular to do with those episodes and has nothing to do with an analogy between private spending on investment and expansionary fiscal policy.