My research considers five broad issues.
1. The theoretical basis of the negative long-run relationship between inflation and the markup
An early overview of the theory can be found in Russell (2006). Note that the empirical work reviewed in this 2006 discussion paper assumes inflation can be approximated as an integrated process of order 1 and this is considered in part 2.1 below. Later work proceeds on the basis of the more realistic assumption that inflation is a stationary process around a frequently shifting mean. This work is considered in part 2.2 below.
The theoretical basis of the negative relationship between inflation and the profitability of firms (i.e. the markup) focuses on the difficulties that firms face when trying to coordinate changes in prices in an inflationary environment. This leads firms to adjust prices cautiously and with a lag to avoid the costs of coordination failure. This results in firms accepting lower profits while adjusting prices. I propose that the uncertainty surrounding the changing of prices is not of the sort that disappears when inflation is stable and that permanently higher inflation will permanently lower the profitability of firms.
The theoretical research also considers the policy implications of the inflation-markup relationship.
2. Estimating the Relationship between Inflation and the Markup
2.1. Assuming Inflation is an Integrated Process of Order 1
The early empirical work identifies the negative relationship between inflation and the markup of price on unit costs (a proxy for profitability) assuming inflation is an I(1) process and the price level is I(2). The research focuses on data for a wide range countries, frequencies, and levels of aggregation. This work demonstrates that it is easy to identify the negative relationship and that the relationship can be characterized as ‘long-run’ in sense of Engle and Granger as the long-run relationship is between two integrated variables of order 1 (i.e. inflation and the markup). Neither variable can be ‘truly’ integrated and only appear to be integrated due to structural breaks in the means of both variables. Consequently, while the ‘true’ statistical process of the variables is most likely stationary around a shifting mean the early empirical analysis proceeds on the assumption that the data can be approximated by an integrated process of order 1. This early empirical research is primarily undertaken with Professor Anindya Banerjee of the University of Birmingham.
2.2. Assuming Inflation is Stationary Process around a Shifting Mean
While inflation is unlikely to be an integrated process as argued in 2.1 above it is likely to have been non-stationary over the past fifty years in the developed world. To argue the converse that inflation is stationary (with a constant mean) implies (i) there has only been one long-run and expected rate of inflation; (ii) one short-run Phillips curve; (iii) the original Phillips (1958) curve did not break down with changes in expected inflation at the end of the 1960s; and (iv) the long-run Phillips curve is a single combination of the long-run rates of inflation and unemployment. Furthermore, if inflation is a stationary process then all the ‘modern’ theories of the Phillips curve since and including F-P are empirically irrelevant as there has been no change in the expected rates of inflation. Unless we are comfortable with these implications of inflation as a stationary process we must conclude that inflation is a non-stationary process.
What then is the likely statistical process of inflation? ‘Modern’ theories of the Phillips curve argue that with no change in monetary policy and with mean zero inflationary shocks then inflation will vary around the long-run rate of inflation. A change in monetary policy results in inflation converging on and varying around a new long-run rate of inflation. We might therefore expect that inflation is a stationary process around a shifting mean. The latter is due to changes in monetary policy and allows for the numerous expected rates of inflation that are central to all the ‘modern’ theories of the Phillips curve since Friedman (1968) and Phelps (1967). If inflation is a stationary process around a shifting mean then the unit root commonly found in inflation data has no behavioural relevance and is simply due to not accounting for the structural breaks in mean inflation when testing for a unit root.
Russell (2011), Russell et al. (2011) and Russell and Chowdhury (2013) estimate Phillips curves using around fifty years of quarterly United States inflation data. They demonstrate that if the shifts in mean inflation are not accounted for in the estimation then the standard results of the ‘modern’ Phillips curve literature can be retrieved. However, once the shifts in mean inflation are allowed for when estimating the models then there is no significant empirical evidence in favour of any of the ‘modern’ theories of the Phillips curve. Importantly, there is also no significant evidence supporting the role of the model defining expected rate of inflation in the NK and hybrid theories of the inflation. It appears that the finding in the standard literature that the dynamic inflation terms sum to one simply reflects the unaccounted shifts in mean inflation when estimating the Phillips curve. This finding is not logically troubling once it is accepted that inflation is a stationary process around a shifting mean as the absolute value of must be less than one by the definition of stationarity.
3. General empirical macroeconomics
The third area of interest is empirical macroeconomics and looks at price and wage inflation, employment, the impact of share prices on the Australian business cycle and the role of exports in transmitting foreign business cycles between countries.
4. Modeling coffee prices
The modeling of coffee prices recognises that changes in government policies over the years leads to changes in the ratio of the producer price of coffee over the terminal price of coffee. Consequently, to model coffee prices successfully using a cointegration analysis or an error correction model requires these changes in the coffee price ratio to be accounted for. If not then the estimated models will be biased and lead to incorrect inference. This modeling approach allows the identification of the long-run producers share of the terminal price of coffee and thereby allows the identification of the losses experienced by producers over the years as a result of different government policies.
5. Methodology: Is Macroeconomics a Science?
When presenting Russell and Chowdhury (2013) at a number of seminars and conferences I was struck by the reaction to the paper by some (many?) economists. These economists would agree with (i) the premise that inflation is a stationary process around a shifting mean, (ii) the empirical work in the paper that accounts for the shifts in mean, (iii) the broad theoretical model in the paper, and (iv) the empirical conclusion that United States inflation over the past fifty years is inconsistent with the New Keynesian, hybrid and Friedman/Phelps expectations augmented theories of the Phillips curve. However, having agreed with all these points in the paper the economists would then state their support for the New Keynesian theory of the Phillips curve because it has representative agent optimising micro-foundations, forward looking ’rational’ agents and rational expectations. This support appeared to me to be highly ‘unscientific’ as it ignores all the empirical evidence that the ‘modern’ theories of the Phillips curve are poor empirical descriptions of the inflation data.
In response Russell (2013) asks what makes a discipline scientific and is macroeconomics a science?
Whether or not macroeconomics is a science depends on the scientific nature of macroeconomic theories and how the discipline responds when the empirical evidence fails to match the underlying assumptions and predictions of the theories. By way of example, Russell (2013) identifies four conditions for macroeconomics to be a science. These conditions identify the types of theories that are relevant for a science and how a discipline uses empirical evidence to examine the theories. The conditions are then applied to the Friedman/Phelps expectations augmented and New Keynesian theories of the Phillips curve as two examples of dominant macroeconomic theories of the past five decades. It is found that while the ‘modern’ theories of the Phillips curve are scientific in the sense of Popper it appears that the empirical validity of the assumptions underlying the model and their associated predictions are somewhat compromised.
Consequently, it is argued that while the discipline in general maintains one condition it routinely violates the other three. This suggests the macroeconomics discipline has some way to go before it can call itself a ‘pure science’.
Friedman, M. (1953). The Methodology of Positive Economics. In M. Friedman (ed.) Essays in Positive Economics, pp. 3-43.
Phelps, E.S. (1967). Phillips curves, expectations of inflation, and optimal unemployment over time, Economica, 34, 3 (August), pp. 254-81.
Phillips, A.W.H. (1958). The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861-1957, Economica, 25, pp 1-17.
Russell, B. (2013) Macroeconomics: Science of Faith Based Discipline?, Dundee Discussion Papers, Department of Economic Studies, University of Dundee, September, No. 276.
Russell, B. (2006). Non-Stationary Inflation and the Markup: an Overview of the Research and some Implications for Policy, Dundee Discussion Papers, Department of Economic Studies, University of Dundee, August, No. 191.
Russell, B., A. Banerjee, I. Malki and N. Ponomareva (2011). A Multiple Break Panel Approach to Estimating United States Phillips Curves, Dundee Discussion Papers, Economic Studies, University of Dundee, June, No. 252.
Russell, B. and R.A. Chowdhury (2013). Estimating United States Phillips Curves with Expectations Consistent with the Statistical Process of Inflation, Journal of Macroeconomics, vol. 35, pp. 24-38.