The Indian inflation–growth relationship revisited: robust evidence from time–frequency analysis

ABSTRACT

This article re-visits the inflation–growth nexus in India using the tools of wavelet, i.e. wavelet correlation, wavelet cross-correlation and scale by scale Granger causality test. Wavelet cross-correlation analysis shows that at the shortest scales inflation and economic growth were independent; at medium scales, there exists feedback effect; and at higher scales, only economic growth is leading to inflation. Furthermore, we find: (a) high and increasing dependence between inflation and economic growth, particularly after mid-2002; (b) high-frequency components of economic growth Granger-cause low-frequency component of CPI-based inflation and vice-versa, and at all scales economic growth Granger-cause inflation at scales of 4–6 and no evidence of causality was detected from WPI-based inflation to economic growth; (c) results indicate that there is no long-run causal link between inflation and economic growth. This study presents new insights for policymakers to sustain economic development by using inflation as an economic tool in India.

KEYWORDS:

• Inflation-growth
• nexus
• time-frequency relationship
• wavelet cross-correlation
• India

JEL CLASSIFICATION:

• C40
• E31
• E32
• E64

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Another line of research estimating the threshold level of inflation. For example, Samantaraya and Prasad (2001) found the threshold level for India to be around 6.5%. Mubarik (2005) found that inflation rate beyond 9% is detrimental for growth in Pakistan and below is favourable. Pattanaik and Nadhanael (2013) examined why persistent high inflation impedes growth in India and identified the factors causing inflation to appear well above the threshold level.

² Shorter widths can introduce undesirable artefacts while wider widths, even if better matching the characteristic feature of the time series, can result in a decrease in the degree of localization of wavelet coefficients.

³ For both spectral and cross-spectra density estimation we lose 20 observations, 10 observations each.

⁴ (j) and (jj) conditions are defined as: (j) $|t_1-t_2|\ge T^{\prime }$and (jj) $|w_1\pm w_2|\ge \pi /h$. For details please refer to Essaadi and Boutahar (2010).

⁵ See Sethi (2012) for additional reasons.

⁶ We following Nachane and Dubey (2011) for the 1992 starting period (see for reason). Eviews-7 and R (2012) was used to analysis.