The stability of the Phillips curve in India: Does the Lucas critique apply?

Abstract

The overwhelming majority of the literature that has attempted to estimate a Phillips curve for India has found that the model does not fit the data, with the notable exception of Paul (2009). However, this paper argues that the Indian Phillips curve can be estimated using standard econometric techniques, as opposed to several special adjustments that are required in Paul (2009)’s work. Once we establish the existence of the Phillips curve in India, we examine whether the Lucas critique applies to the behavior of the model under different Reserve Bank of India (RBI) Governors. We find that there is little-to-no evidence for structural change in the model between successive RBI Governors. Hence the results demonstrate that the Phillips curve is alive and well in India, and that the Lucas critique does not apply empirically to Indian inflation–output data.

Introduction

The Phillips curve, which famously states that a negative relationship exists between inflation and unemployment¹ in the short-run, does not apply in India according to the majority of the existing literature. There are several examples that one could name of researchers who argue that the Phillips curve is nonexistent in India. For instance, Bhalla (1981) argues that India’s inflation rate responds mainly to changes in the money supply as opposed to a measure of real economic activity, while Ghani (1991) and Virmani (2004) find a negative relationship between output and inflation, which is counter to the predictions of the Phillips curve.

While these studies may dismiss the applicability of the Phillips curve to Indian data, the search for the existence of the model is still important since the negative relationship between inflation and unemployment (or a positive relationship between inflation and output) remains important from the policymaker’s standpoint. Indeed the nonexistence of the Phillips curve would make policy choices less clear, not to mention that this notion contradicts the experience of most other nations. In particular, the empirical accelerationist Phillips curve has been quite robust to U.S. data (see Fuhrer, 1995) as well as to other developed economies (for example, see Paloviita, 2005 for recent empirical evidence on the Phillips curve in Europe), so it seems logical that the model should be applicable to other countries as well.

Paul (2009) estimates a Phillips curve for India in fitting with Gordon (1982)’s ‘triangle’ model, and finds that the inclusion of supply shocks does in fact make a Phillips curve appear in Indian inflation data. However, Paul is forced to make some unusual adjustments to his dataset in order to find evidence of the model’s applicability to the data. Specifically, Paul finds that the short-run tradeoff between inflation and output only emerges if we examine the crop year instead of the fiscal year. Moreover as part of his empirical investigation, Paul uses annual data which precludes a serious investigation on the autoregressive dynamics of inflation. Further still, when it comes to incorporating supply shocks into the model, the author uses a dummy variable to account for oil price shocks rather than using actual oil price data. While the overall result of Paul (2009)’s paper is extremely interesting, this paper argues that the method by which we get to those results can be greatly improved in a manner that is more intuitive and more in keeping with how Phillips curves are estimated for developed countries.

In particular, this paper improves upon the existing literature that estimates a Phillips curve for India in several ways. First, we use quarterly data which also span a longer time period than most of the existing literature does; namely this paper examines the inflation–output tradeoff from 1970 to 2008. Second, the use of quarterly data over a longer time horizon allows us to introduce autoregressive dynamics of inflation to our Phillips curve model. This is in keeping with the popular idea that inflation tends to be highly persistent (Fuhrer & Moore, 1995). In addition, this paper does this by a typical optimal lag length selection process rather than imposing an arbitrary number of lags of the dependent variable to the model. Third, this paper addresses whether nonstationarity exists amongst the data. Finally, this paper also finds no need to use the crop year to separate the data as Paul (2009) does, which allows us to use standard econometric techniques to estimate the Phillips curve for India.

After establishing the existence of a Phillips curve for India, this paper then examines the stability of the model in a similar way as Fuhrer (1995) does using U.S. data. Namely, Fuhrer identifies that one of the main criticisms aimed at the Phillips curve is that the model is not structural. This is in essence what the Lucas (1976) critique argues: that the Phillips curve will not be stable over long periods of time. In other words, as monetary policy changes there is no theoretical reason as to why the parameters underlying the Phillips curve will remain the same. Moreover, even if the underlying behavior of economic agents remained stable between different policymaking regimes, the relationships between aggregate variables could easily change as conditions varied in the linkages that are not captured by the aggregate relationships. However, Lucas and Sargent (1978) argue that whether this critique applies to models such as the Phillips curve or not is ultimately an empirical matter and not a theoretical one. Hence, Fuhrer (1995) investigates the performance of the U.S. Phillips curve from the 1960s until the 1990s and finds that the Phillips curve is highly robust over this time period and concludes that the Lucas critique is not empirically applicable to the U.S. post-war Phillips curve. In this paper we investigate a similar question by examining whether the Phillips curve has been stable between different monetary policy-setting regimes in India.

Overall, this paper finds that the short-run tradeoff does indeed exist in India. In other words, contrary to the majority of the existing literature, the Indian Phillips curve is “alive and well”. Indeed this positive relationship between inflation and the output gap can be clearly discerned in India’s data for the last 38 years. Moreover, this paper finds that there is also little evidence of Phillips curve instability in India from 1970 until 2008. This leads us to conclude that the Lucas critique does not apply to the Indian inflation–output short-run tradeoff, at least from an empirical point of view. Hence not only does the Phillips curve exist in India, it is also a highly robust aggregate relationship that has been valid for almost 40 years.

Section 2 of the paper reviews and examines the findings of the existing literature, while Section 3 discusses how we model and estimate the Phillips curve for India. Section 4 investigates the empirical applicability of the Lucas critique to the model, and Section 5 summarizes our findings.