João Madeira discusses his research in this area.

Economics undergraduate students worldwide are taught business cycle fluctuations using the IS-LM model, the AD-AS model, and the Phillips curve. The key message from these models is that monetary and fiscal policy can be effective in stimulating aggregate economic output in the short to medium run (before prices of goods and services have adjusted) but not in the long run (because prices adjust). This key message remains valid in the New Keynesian models which have become the most widely used framework to understand business cycles in graduate/doctorate programmes, academic research, and also in central banking. These are models which, unlike those taught in undergraduate programmes, are: 1) dynamic, which allows to study how the economy changes over time; and 2) based on optimising the decision-making of households/ firms, and which therefore take into account that agents adjust their behaviour in response to changes in economic policy.

Although New Keynesian models have greatly enhanced our understanding of inflation and aggregate output dynamics, they have nonetheless been subject to criticism. In particular, several researchers pointed out that: 1) regression estimates rejected a statistical significant relationship between inflation and the cyclical level of aggregate economic activity, as predicted by the New Keynesian theory; 2) the most widely used measure of economic activity is countercyclical (which is not plausible, since in periods of economic expansion one would expect that workers require a higher real wage in order to be induced to supply extra hours); and 3) given the estimated sluggishness of inflation responses to changes in output, the standard New Keynesian model implies that prices adjust on average only every 18 to 24 months (the micro data indicates that prices adjust on average every 4 to 5 months).

In my work I show how a simple modification to the standard New Keynesian model allows for all the above criticisms to be addressed. In the standard model, workers can instantly move without costs between firms. This assumption makes the model simpler to solve. I study what happens in the more realistic case when it involves time (and costs) for firms to change the numbers of their workforce.

One implication of this change is how best to measure the cyclical level of aggregate economic activity. If it takes time for firms to hire new workers then in order to increase output in the short run the firm must make its current staff work more hours. This indicates that a good measure of the cyclical level of aggregate economic activity should be based on costs with overtime labour.

My research constructs such a measure using data from the US Bureau of Labor Statistics (BLS) on the number of persons who worked 41 hours and over per week. I then show that regression estimates of a New Keynesian model that allows for forward-looking and backward-looking expectations show a statistically significant relationship between inflation and the cyclical level of aggregate economic activity (when measured using costs with overtime labour), as predicted by the theory. Moreover, the cyclical level of aggregate economic activity when measured with overtime labour costs was procyclical (as is plausible, if demand shocks account for a relevant component of the business cycle).

Finally, the New Keynesian model that I have developed implies that prices adjust on average every 4 to 5 months (as indicated in the micro data). The intuition is as follows. A model in which firms cannot adjust their number of workers instantly implies that a firm’s marginal cost is no longer independent of its own level of output (whereas in the standard model, a firm’s marginal costs are constant regardless of how much output is produced, which is not very realistic). In this case, a firm that contemplates raising its price understands that this implies less demand and therefore less output. The reduced output implies a lower marginal cost. Other things being equal, the lower marginal cost induces a profit maximising firm to post a lower price. Therefore, in my model price-adjusting firms are induced to keep their relative price close to the non-adjusters. Hence, the sluggishness of inflationary responses to changes in output can be reconciled with individual firms’ flexibility in changing prices.

Essentially the New Keynesian model that I have developed implies that firms make frequent price adjustments, but that these are small in magnitude. In other words, frequent price adjustment does not mean that prices adjust fully to economic shocks in a short period of time. This allows monetary and fiscal to be effective in stimulating output in the short to medium run even though average prices do not stay unchanged for long.

In subsequent research I have shown that taking into account the fact that workers cannot instantly move between firms without costs allows the New Keynesian model to fit better with aggregate time series data (consumption, investment, GDP growth, wages, employment numbers, inflation, and interest rates) This is an important contribution because models which fit the data better can be trusted to give more reliable forecasts to economic policy changes.