Inflation-Growth Relationship: New Evidence for India

Abstract

This paper revisits the threshold level of inflation for India. The empirical analysis follows spline regression for the period 1996–97Q1 to 2019–20Q4. The results suggest the existence of a statistically significant threshold level of inflation at 5 to 5.5% in terms of both CPI and WPI. Below this level, the impact of inflation on growth is generally positive whereas it is negative above this level, and therefore injurious to growth. Hence, policymakers in India may consider reducing the tolerable band from 4 ± 2% to 4 ± 1.5% under the flexible inflation targeting regime with a vision to further compress it to 4 ± 1% in due course.

Introduction

The main objective of public policy in most countries has been to achieve high and sustainable GDP growth with low inflation consistent with steady-state growth. Up to the late 1960s, the relationship between inflation and growth was either non-existent or believed to be marginally positive. The persistence of high inflation and low growth in the 1970s changed the whole scenario leading to empirical investigation on the precise relationship between inflation and economic growth. Most empirical works suggest that inflation affects growth positively up to a specific 'threshold' and harms thereafter (Fischer 1993; Sarel 1996; Barro 1995; Ghosh and Phillips 1998; Khan and Senhadji 2001, etc.). High inflation creates uncertainty affecting real savings, real investment, and exchange rates, which leads to the misallocation of resources that impedes growth. The threshold inflation at which growth is maximised is generally higher for developing countries than for developed and industrial countries (Khan and Senhadji 2001; Jha and Dang 2012; Espinoza et al. 2012; Kremer et al. 2013). The threshold level of inflation is a dynamic concept, that changes over time depending on several factors such as the structural transformation of the economy, degree of cross-border openness, financial innovations, regime change in the framework of monetary policy, and above all variations in total productivity (Khan and Senhadji 2001; Vinayagathasan 2013; Pattanaik and Nadhanael 2013), etc.

Being an emerging market economy, India witnessed momentous changes on multiple fronts since the early 1990s. Notable among them are the structural transformation of the economy, increasing openness, financial sector reforms, and new frameworks of monetary policy from broad monetary targeting in the mid-1980s to multiple indicator approach in the late 1990s and to flexible inflation targeting (FIT) in 2016. In response, the economy registered a higher GDP growth during the last three decades compared to the previous ones. In India, several studies have examined the inflation-growth relationship mostly in the context of a closed economy. Previous studies did not include the period after the introduction of FIT. Moreover, the RBI has switched over to CPI inflation as a nominal anchor under the FIT regime. Although several studies have found the existence of a ‘threshold’ level of inflation in India, due to the structural transformation of the economy and regime change in the framework of monetary policy, the level of threshold inflation might have changed over time.

The objectives of this paper are (a) to test the causality between inflation and growth; (b) to revisit the threshold level of inflation in India; (c) to use both CPI and WPI inflation in an open economy framework to investigate the probable change in the threshold level of inflation in India, and (d) draw policy implications in light of new findings. The rest of the paper is organized as follows: the section "Review of Literature" reviews the relevant literature; the section "Data and Methodology" presents the data and methodology used; the section "Analysis of Empirical Results" analyses the empirical results, including the robustness of results through suitable diagnostic tests; and the section "Concluding Observations and Policy Implications" provides concluding observations and policy implications.

Review of Literature

There is a large body of literature, which examined the precise relationship between inflation and growth in both developed and developing countries. The findings of these studies are mixed and often contradictory. First, there is no relationship between growth and inflation (Sidrauski 1967; Cameron et al 1996). Second, there is a long-run positive relationship between growth and inflation (Mundell 1963; Mallik and Chowdhury 2001; Behera 2014; Behera and Mishra 2016). Third, there is a short-run negative relationship between growth and inflation (Bhatia 1960; Faria and Carneiro 2001; Erbaykal and Okuyan 2008; Bhaduri 2016). Fourth, there is a non-linear relationship between growth and inflation—inflation affects growth positively below some critical level or point of inflexion/threshold, and negatively above that level (Fischer 1993; Sarel 1996; Ghosh and Phillips 1998; Khan and Senhadji 2001; Bick 2010; Vinayagathasan 2013; Mohaddes and Raissi 2014; Thanh 2015; Aydin et al. 2016), etc.

The seminal work by Fischer (1993) examined for the first time the possibility of a non-linear relationship between inflation and economic growth. Using both cross-section and panel data for a sample of ninety-three countries that included both developing and industrialized countries, he found the presence of non-linearities in the relationship between inflation and growth. Barro (1995) studied the relationship using data from over 100 countries from 1960 to 1990 and found threshold inflation at 10%. Based on annual data from 1970 to 1990 for 87 countries including India, Sarel (1996) found a significant structural break in the inflation and growth relationship around 8% inflation. Inflation above 8% harms growth and did not have any effect on growth, or it was slightly positive below 8%. Ghosh and Phillips (1998) used panel regressions allowing for a nonlinear specification covering IMF member countries during 1960–1996 and found evidence of a lower threshold of 2.5% while acknowledging that 5% or 10% thresholds work almost as well statistically. Khan and Senhadji (2001) used unbalanced panel data with 140 countries comprising both industrial and developing countries for a period from 1960 to 1998 and found that the threshold level of inflation above which it significantly slowed growth was estimated at 1–3% for industrial countries and 11–12% for developing countries. By dividing the sample of 80 countries over the 1961–2000 period by decades, Pollin and Zhu (2006) found that the threshold inflation would be around 15–18%. Bick (2010) using Hansen's (1999) panel threshold model with regime intercepts found the threshold at 12% for 40 developing countries from 1960 to 2004. Jha and Dang (2012) examined the impact of inflation variability on economic growth and found that in developing countries when the rate of inflation exceeded 10% it adversely impacted growth. In a study based on a large panel dataset including 124 countries for the period 1950–2004, Kremer et al. (2013) estimated that inflation exceeding 17% was associated with lower economic growth for developing countries and 2.53% for industrialized countries.

Das and Loxley (2015) examined the non-linear relationship between inflation and economic growth for 54 developing countries over the 1971–2010 period, using Pollin and Zhu (2006) technique. They found the threshold rate of inflation at 23.5% for Latin America and the Caribbean, approximately 11% for Asia, and 23.6% for Sub-Saharan Africa. Thanh (2015) using the Panel Smooth Transition Regression (PSTR) model over the period 1980–2011, found a threshold level of inflation at 7.84%, above which inflation started impeding economic growth in the ASEAN-5 countries, namely Indonesia, Malaysia, the Philippines, Thailand, and Vietnam. Through dynamic panel data analysis, Aydin et al. (2016) investigated the influence of inflation on economic growth for five Turkish republics (Azerbaijan, Kazakhstan, Kyrgyzstan, Uzbekistan, and Turkmenistan) that were in transition and found a nonlinear relationship between inflation and growth with the threshold at 7.97%.

In the Indian context, the Chakravarty Committee (1985) suggested that the acceptable level of inflation in India is 4%. Vasudevan et al. (1998) used monthly data on IIP from April 1976 to March 1998 and annual data on GDP from 1961 to 1998 and found the threshold level of inflation at 6–6.5%. The results of Kannan and Joshi (1998), Pattanaik and Nadhanael (2013), and Singh (2010) are also on a similar line as they found the threshold level at and/or above 6%. In contrast, Singh and Kalirajan (2003), using spline regression and annual data for the period 1971–1998, could not find any threshold level of inflation for India. According to them, inflation rising from any level harms economic growth. The RBI Annual Report 2010–11 estimated threshold inflation for India in the range of 4–6% using three alternative techniques i.e., spline regression, non-linear least squares (NLLS), and Logistic Smooth Transition Regression (LSTR) models. Based on the quarterly data for the period Q1:1996–97 to Q3: 2010–11, Mohanty et al. (2011) and Mohaddes and Raissi (2014) examined the growth-inflation nexus in India and empirically investigated the existence of any threshold level of inflation using different methodologies and found that there was a statistically significant structural break in the relationship between output growth and inflation between 4.0 and 5.5% inflation. Using both WPI and CPI-C (rural–urban combined), the RBI (2014) rediscovered the threshold at 6.7% and 5.8% for CPI-C and WPI inflation respectively. Behera and Mishra (2017) by employing the spline regression method for the period 1990–2013, found a structural break in the relationship between inflation and economic growth at 4% beyond which inflation negatively affected economic growth. Based on the macro-theoretic model, Dholakia (2020), using CPI inflation and annual data from 1995–96 to 2017–18, found threshold inflation at 5.4–6.0%.

The recent findings show somewhat lower threshold inflation for India than the earlier findings. Most Indian studies were based on WPI inflation in a closed economy framework as the long-time series on CPI was not available on a pan-India basis. In the context of structural transformation, a periodic revisit of the issue is required as the change in threshold inflation has significant policy implications.

Data and Methodology

This study used quarterly data obtained from the Handbook of Statistics on the Indian Economy, published by RBI. It covers a period from 1996–97Q1 to 2019–20Q4, as India’s quarterly data on GDP is not available before 1996–97. Data for the next two years are excluded from the sample as these are highly distorted due to the COVID-19 pandemic. The dependent variable is the real GDP growth represented by the growth of gross value added at the basic price (Base year: 2011–12 = 100). The inflation rate which is the percentage change in CPI-C on a point-to-point basis is the major explanatory variable. For control variables, the real effective exchange rate of the Indian rupee (trade-weighted, Base: 2015–16 = 100), and lagged values of GDP growth are used. WPI inflation is also examined to compare the results. Data are seasonally adjusted. Since GVA data are not available before 2011–12Q1, GDP at factor cost is converted to GVA using the splicing method. Similarly, CPI-IW data are converted to CPI-C data for the period 1996–97Q1 to 2011–12Q4 using the splicing method.

The results derived from the regression models are likely to be spurious if data are non-stationary. Therefore, variables are tested for stationarity using Augmented Dickey–Fuller (ADF) and Phillips Perron (PP) unit root tests.

The Granger causality test (Granger 1969 and Sims 1972) is used to ascertain the direction of causality between inflation and growth in India. Two null hypotheses; ‘growth does not Granger cause inflation’ and ‘inflation does not Granger cause growth’ are tested for both CPI and WPI inflation. The available results of causality can be divided into two categories. First, there is a unidirectional causality that runs from inflation to growth and vice versa (Paul et al. 1997; Mubarik 2005; Fabayo and Ajilore 2006; Erbaykal and Okuyan 2008; Bhaduri 2016; Behera and Mishra 2017). Second, there is bidirectional causality between inflation and growth (Paul et al. 1997; Behera 2014; Behera and Mishra 2016). However, the causality tests are sensitive to country-specific studies covering different periods. Most countries have mixed results.

To estimate the level of inflation threshold at which growth is maximised, the study used the spline regression method, suggested by Sarel (1996) and Khan and Senhadji (2001). The specification of the model is as follows:

$${\text{Y}}_{{\text{t}}} = \, \beta_{0} + \, \beta_{{1}} \pi_{{\text{t}}} + \, \beta_{{2}} {\text{D}}_{{1}} \left( {\pi_{{\text{t}}} {-} \, \pi_{{\text{t}}}^{*} } \right) \, + \, \beta_{{3}} {\text{ER }} + \, \beta_{{4}} {\text{Y}}_{{{\text{t}} - {1}}} + {\text{ e}}_{{\text{t}}}$$

(1)

where:

Y_t = Real GDP growth.

π_t = Inflation rate (both CPI and WPI).

π^* = Assumed threshold inflation.

ER = Real effective exchange rate (REER).

D₁ = Dummy variable for inflation below and above the threshold:

D₁ = 1 if π_t > π^*

 = 0 if π_t ≤ π^*

e_t = Error term.

Since π* is unknown, the OLS regression is iterated with different π* values. The optimal π* is obtained from the regression that minimizes the residual sum of squares (RSS) or maximizes the adjusted R-square. β₁ shows the impact of inflation on growth and β₂ indicates the impact of inflation on growth above the threshold inflation. A priory, one may expect β₁ to be positive and β₂ negative. The lagged dependent variable is presumed to be determined by a vector of variables that normally influence growth on which there is no consensus. Inflation and its threshold are unlikely to influence the lagged dependent variable. As the dummy is zero, while β₁ is relevant up to the threshold, the sum of β₁ and β₂ is appropriate beyond the threshold. These statistics would be examined from the regression results. A few diagnostic tests would also be conducted on the preferred regressions that determine India’s threshold level of inflation.

Analysis of Empirical Results

Unit Root Tests

Variables used in the study do not have unit roots as is evident from Table 1. The ADF and PP tests are based on the null hypothesis that the variable has a unit root, which has been rejected.

Table 1 Unit Root Test

Granger Causality Test

The Granger Causality test is applied to measure the linear causation between inflation and economic growth. In the case of CPI inflation, Table 2 shows that the null hypothesis—‘inflation does not Granger cause growth’ is rejected at a 5% level of significance, which means that CPI inflation significantly impacts GDP growth in India. The second null hypothesis—‘growth does not Granger cause inflation’ is not rejected. Hence, the causality between CPI inflation and growth is unidirectional.

Table 2 Granger Causality Test (CPI Inflation)

In the case of WPI inflation, Table 3 shows that the null hypothesis—‘inflation does not Granger cause growth’ is rejected at a 10% significance level. The second null hypothesis—‘growth does not granger cause inflation’ is also rejected at a 5% significance level. Hence, the causality between WPI inflation and growth is bi-directional. Growth is essentially a supply-side variable although it has a corresponding demand side too. India’s WPI is roughly close to producers’ price index and therefore can be interpreted as a supply-side variable too. Hence, the bidirectional causality between two supply-side variables -WPI inflation and growth—is not surprising. There might be a simultaneity problem between growth and WPI inflation in India similar to many other countries, which is beyond the scope of this study.

Table 3 Granger Causality Test (WPI Inflation)

Estimation of India’s Inflation Threshold

The search for the optimal threshold level of inflation is done by estimating Eq. (1) for π^* = 2.5% to 10%, with an increment of 0.5% at a time. The results are presented in Tables 4 and 5 for CPI and WPI inflation respectively. In the case of CPI inflation (Table 4), it can be observed that both inflation and inflation dummy have a significant influence over growth between 4.5% and 7.5% threshold levels of inflation. The exchange rate is not significant and its impact on growth looks innocuous given the low passthrough of the exchange rate to GDP. Other external sector variables such as the change in oil prices and the current account deficit as a proportion to GDP were also tried but the coefficients (not reported here) were insignificant. These variables might be influencing GDP at appropriate lags captured in the lagged dependent variable. At the 5% threshold, adjusted R² is maximum i.e., 0.7308. At 5.5%, the adjusted R² is more or less the same i.e., 0.7307. In both cases (highlighted in Table 4), β₁ is positive and β₂ is negative and both are statistically significant at a 5% level. The sum of the coefficients of inflation and the dummy turns out to be negative supporting our hypothesis that inflation above the threshold harms growth. Hence, we believe that 5–5.5% could be the new threshold level of inflation in India compared to the 6–6.5% found earlier.

Table 4 Search for Threshold Using CPI Inflation

Table 5 Search for Threshold using WPI Inflation

In the case of WPI inflation (Table 5), although inflation is not significant, the dummy is significant around the threshold level of inflation. The exchange rate is insignificant similar to the regression results relating to CPI inflation. At the 5.5% threshold (highlighted in Table 5), the adjusted R² is maximum i.e., 0.5511. The signs of β₁ (between 3.5% and 6%) and β₂ are positive and negative respectively as expected. Like CPI inflation, the sum of the coefficients of WPI inflation and its threshold is negative. Hence, 5.5% inflation is also the new threshold level measured in WPI. However, the adjusted R² relating to WPI regression looks slightly weaker than that of CPI inflation. The implication of India’s new threshold level of inflation from 5 to 5.5% would be reflected in the concluding section.

The possibility of very low inflation harming growth cannot be ruled out. At the 2.5% threshold, the coefficients of both CPI and WPI inflation are negative (Tables 4 and 5). As the dummy variable is zero at this level, its coefficient is not meaningful. India’s average inflation coming down to 2.5% could be as harmful to growth as high inflation above 5.5%.

The adjusted R² is plotted against different levels of threshold in Figs. 1 and 2, for CPI and WPI inflation respectively. The visual representations explain the story loud and clear.

Fig. 1

Source: Authors’ Estimation as in Table 4

Threshold Inflation Measured by CPI.

Fig. 2

Source: Authors’ Estimation as in Table 5

Threshold Inflation Measured by WPI.

The coefficient of inflation (β₁) up to the threshold and the sum of coefficients of inflation and Dummy (β₁ + β₂) beyond the threshold from Tables 4 and 5 are presented in Figs. 3 and 4, respectively. At very low WPI and CPI inflation, around 2.5%, β₁ is negative. At the threshold level of inflation and above, β₁ + β₂ is also negative. The positive impact of the CPI inflation on growth is clearly visible between 3% and 5%, but it is not so pronounced for the WPI inflation, possibly due to bidirectional causality between WPI inflation and growth. Before suggesting policy implications, it is necessary to check the robustness of the empirical results presented above.

Fig. 3

Source: Table 4

Impact of CPI Inflation on Growth up to the Threshold and Above.

Fig. 4

Source: Table 5

Impact of WPI Inflation on Growth up to the Threshold and Above.

Robustness of Threshold Inflation

Our preferred regressions for CPI and WPI, chosen from Tables 4 and 5, are given in Table 6.

Table 6 Preferred Regression Results

If CPI inflation is around the threshold of 5%, growth is estimated to increase by 0.33% and if it is above that then growth is estimated to decrease by 0.39%. In the case of WPI inflation, the positive effect of inflation on growth is very small and insignificant, but the dummy is negative (– 33%) and significant. The sum of the coefficients of inflation and the dummy in the case of both CPI and WPI turns out to be negative supporting our hypothesis that inflation above the threshold harms growth.

Durbin’s ‘h’ statistics are well below the critical value and therefore there is no first-order autocorrelation. Additional diagnostic tests done here relate to the Jarque–Bera normality test, Breusch–Godfrey LM Test for serial correlation, Breusch–Pagan–Godfrey (BPG) test for heteroscedasticity, and CUSUM and CUSUM of squares tests for stability of the results (details in Annexe 1).

The JB test conducted to test the normality of the residuals could not reject the null hypothesis that residuals were normally distributed (Tables 7 and 8). Similarly. the Breusch–Godfrey LM test for serial correlation was conducted to check the first-order autocorrelation in the models. The null hypothesis of no serial correlation could not be rejected. The Breusch–Pagan–Godfrey (BPG) test conducted to check for heteroscedasticity also failed to reject the null hypothesis of homoscedasticity.

Table 7 Diagnostic Test—Threshold Regression (CPI Inflation)

Table 8 Diagnostic Test—Threshold Regression (WPI Inflation)

To check the stability of parameters/regression results, CUSUM and CUSUM of squares tests were conducted. Figures 5, 6, 7, 8 show that the estimated models are stable as both test statistics are within the 5% corridor.

Fig. 5

CUSUM Test – Stability of Parameters (CPI Regression)

Fig. 6

CUSUM of Square Test – Stability of Parameters (CPI Regression)

Fig. 7

CUSUM Test – Stability of Parameters (WPI Regression)

Fig. 8

CUSUM of Square Test – Stability of Parameters (WPI Regression)

Concluding Observations and Policy Implications

The positive trade-off between growth and inflation has been intensely debated since the early 1970s. Inflation does not ‘grease the wheel of commerce’ perpetually. Inflation may put ‘sand on the wheel of commerce’ at a high level of inflation that varies among countries. The non-linear relationship between growth and inflation has been empirically established in the literature. In the case of developed countries, the threshold level of inflation at which the positive effect of inflation on growth is maximized remains at a relatively low level compared to the developing countries. This has significantly influenced policymakers to decide on the numerical inflation target and/or the tolerance band.

In India too, under the flexible inflation targeting since 2016, the government has set the CPI inflation target at 4% with a wide tolerance band of ± 2%. This was based on the empirical studies available at that time which believed that the inflation-growth relationship in India was non-linear and the threshold inflation seemed to lie between 6 and 6.5%. We have also got a non-linear relationship between growth and inflation in India. However, the point of inflexion seems to have changed from 6–6.5% to 5–5.5% in terms of both CPI and WPI inflation. Because of new findings, there is a case for reducing the inflation tolerance band from 4 ± 2% to 4 ± 1.5% in India at the earliest. As India should play a dominant role in the global economy going forward for which price stability is crucial, endeavours should be made to target CPI inflation at 4% with a target band of ± 1% in the medium term.

Appendices

Annexe 1: Diagnostic Tests

To check the robustness of the regression the following diagnostic tests are done.

Jarque–Bera Normality Test

Jarque–Bera (JB) Test, which is used to test the normality of the data and residuals is an asymptotic test. The test statistic is:

$$JB=n \left[\frac{{S}^{2}}{6}+ \frac{{\left(K-3\right)}^{2}}{24}\right]$$

where, $n$ = Sample size, $S$ = Skewness coefficient, and $K$ = Kurtosis coefficient. It is used to test the joint hypothesis that $S$ = 0 and $K$ = 3 (JB statistic is expected to be zero). If the p-value is reasonably high, the null hypothesis cannot be rejected which means that the variable is normally distributed.

Breusch–Godfrey LM Test for Serial Correlation

Breusch–Godfrey (BG) LM test is used in this study to check the autocorrelation in residuals besides Durbin’s h statistics. This test estimates the following regression:

$${\widehat{u}}_{t}= {\alpha }_{1}+{\alpha }_{2} {X}_{1t}+ {\alpha }_{3} {X}_{2t}+. . . +{\alpha }_{k} {X}_{kt}+{\widehat{\rho }}_{1}{\widehat{u}}_{t-1}+{\widehat{\rho }}_{1}{\widehat{u}}_{t-2}+\dots +{\widehat{\rho }}_{p}{\widehat{u}}_{t-p}+{\varepsilon }_{t}$$

where ${\widehat{u}}_{t}$ = error term of the original regression, ${X}_{it}$= independent variables. The null hypothesis (there is no serial correlation) is as follows:

$${H}_{0}= {\rho }_{1}={\rho }_{2}=\dots ={\rho }_{p}=0$$

If the sample size is large, the test statistic is,

$$(n-p){R}^{2}\sim {\chi }_{p}^{2}$$

If it does not exceed the critical value then the null hypothesis is accepted that there is no serial correlation.

Durbin’s ‘h’ Statistic

Durbin-Watson’s ‘d’ statistic is used to test the presence of serial correlation in the error term. When the model includes lagged dependent variable Durbin’s ‘h’ statistic is used to test for first-order autocorrelation. The h statistic is defined as:

$$h= \left(1-\frac{d}{2}\right) \sqrt{\frac{n}{1-n\left[var \left(\widehat{\beta }\right)\right]}}$$

where, n = sample size, $var \left(\widehat{\beta }\right)$ = variance of coefficient of the lagged dependent variable, d = Durbin-Watson d statistic. If the sample size is large and computed $\left|h\right|$ > 1.96, then the null hypothesis (no serial correlation) is rejected and there exists a serial correlation.

Breusch–Pagan–Godfrey (BPG) Test

Breusch–Pagan–Godfrey (BPG) Test is used to test whether residuals are homoscedastic. This test is based on the following regression:

$${\rho }_{i}={\alpha }_{1}+ {\alpha }_{2}{Z}_{2i}+\dots + {\alpha }_{m}{Z}_{mi}+{v}_{i}$$

where ${\rho }_{i}$ is the ratio of squared residuals and error variance of the original model and ${v}_{i}$ is the error term.

The test statistics is:

$$\Theta =\frac{{\text{ESS}}}{2} \sim {\chi }_{m-1}^{2}$$

ESS is the explained sum of squares of the above regression. If $\Theta$ exceeds the critical value, one can reject the null hypothesis of homoscedasticity.

CUSUM and CUSUM of Squares Tests

The cumulative sum (CUSUM) of recursive residuals and CUSUM of squares (CUSUMSQ) tests are conducted to check the stability of the estimated parameters. CUSUM and CUSUM of squares detect the unexplained systematic change and sudden change in coefficients respectively. If values of the test statistics lie outside an acceptable range, then it rejects the null hypothesis of stability of the coefficients.