ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Examining the Decoupling Hypothesis for India

This paper examines the decoupling hypothesis for India. It analyses business cycle synchronisation between India and a set of industrial economies, particularly the United States, over the period 1992 to 2008. The evidence suggests that the Indian business cycle exhibits increasing co-movement with business cycles in industrial economies over this period. Indian business cycle synchronisation is stronger with industrial countries as a whole as opposed to the co-movement found with the us.


Examining the Decoupling Hypothesis for India

Shruthi Jayaram, Ila Patnaik, Ajay Shah

This paper examines the decoupling hypothesis for India. It analyses business cycle synchronisation between India and a set of industrial economies, particularly the United States, over the period 1992 to 2008. The evidence suggests that the Indian business cycle exhibits increasing co-movement with business cycles in industrial economies over this period. Indian business cycle synchronisation is stronger with industrial countries as a whole as opposed to the co-movement found with the US.

This paper was written under the aegis of the SPF Financial and Monetary Policy Reform Project at the National Institute for Public Finance and Policy, New Delhi. We gratefully acknowledge help from Matthieu Stigler, Giovanni Veronese, Eric Dubois, Charles Adams, Radhika Pandey and Rudrani Bhattacharya. The paper has been implemented using open source software: R Scilab, GROCER and X12-Arima. All code is available on request from The authors are grateful to the comments of a referee of this journal.

Shruthi Jayaram (, Ila Patnaik (, Ajay Shah ( are with the National Institute for Public Finance and Policy, New Delhi.

Economic & Political Weekly

october 31, 2009 vol xliv no 44

1 Introduction

ndia has seen greater integration with the world economy through trade in goods and services, and through financial integration over the past two decades. Has this integration been accompanied by business cycle synchronisation with the rest of the world? Or, is India in a period of high economic growth that is decoupled from the rest of the world?

The literature on developed countries suggests that increasing trade intensity leads to business cycle synchronisation, but there is no consensus, either in the theory or in the evidence, on what might come about with developing economies. This has given rise to the debate about a possible “decoupling” of the business cycle in emerging markets, especially in India and China, from that found in developed countries. The apparent divergence in the performance of different regions of the world economy in 2008 brought the theme of decoupling to the forefront of debates on the international economy (Kohn 2008).

The early literature, which focused on developed countries, found ample evidence that increasing trade intensity leads to increased business cycle synchronisation (Frankel and Rose 1998). More recent work on emerging markets shows mixed results, with Agenor et al (2000) and Calderon et al (2007) finding an increase in output correlations over time and Fidrmuc et al (2008) finding evidence of decoupling. Chan and Khong (2007) find that Asia-Pacific economies tend to be more correlated with Japan than the US. Studies such as Kose et al (2003) find that increased trade and financial liberalisation adds to contagion of macroeconomic and trade shocks. The findings of Kose et al (2008) seem to suggest evidence in favour of decoupling between industrial countries and emerging economies.

Disagreements in the empirical literature arise from the differences in countries and time periods studied, alternative detrending techniques and business cycle “identification” procedures, accounting for production asymmetries and the impact of interindustry trade (specialisation and divergence) versus that of intra-industry trade (common shocks and convergence) on the business cycle (Kose and Yi 2001; Frankel and Rose 1998; Krugman 1993). Cycle interactions between different economies also pose challenges, with shocks precipitating and dissipating via feedback loops across various countries. Data availability, changes in the policy environment and structural breaks in trend growth are some more of the accepted difficulties of estimating business cycle synchronisation in emerging markets.

While anecdotal evidence for India suggests increased linkages with the world, the systematic evidence on this is limited. India is part of the sample of countries studied by Agenor et al (2000) and Fidrmuc et al (2008). The latter paper examines the case of India and China and finds evidence in favour of decoupling. Similarly, Akin and Kose (2008) find that countries of the “Emerging South” (that includes India and China) have decoupled with industrial countries over time.

Another dimension of exploration lies in linkages with the US as opposed to other industrial countries. The US has strong trade and financial links with India.1 In addition, Indian monetary policy has often consisted of a de facto pegged exchange rate, which generates a channel for transmission of US monetary policy into the Indian economy (Patnaik 2007). Hence, it is useful to measure the extent to which the Indian business cycle is synchronised with the US, as opposed to synchronisation with a broader set of industrial countries.

In this paper, we use output and trade data on India and the rest of the world to investigate three questions: (1) How has the Indian business cycle behaved during world expansions and recessions?

(2) Has there been a change in business cycle synchronisation over time between India and the rest of the world? (3) Does India have particularly strong linkages with the US, or is the co-movement stronger with a broad set of industrial countries?

It can be seen that these questions are only of correlation and not causation. In the context that there is no consensus in the literature on the impact of increasing trade and financial liberalisation on business cycle integration; establishing or rejecting the synchronisation hypothesis is in itself an important element in the debate. The sharp increase in economic integration between India and the rest of the world across 1992-2008 suggests that business cycle synchronisation could have changed over this period, necessitating its separate measurement.

We construct a data set consisting of measures of industrial production for India and advanced economies and a coincident indicator for the US business cycle. In addition to exploratory data analysis, we use the Harding-Pagan index of concordance to measure the extent of synchronisation.

Our results show that the Indian business cycle is linked to business cycle conditions in the US and the rest of the world with statistical and economic significance. We find that there is an increase in this synchronisation over the period 1992-2008. Finally, we find that the Indian business cycle is more synchronised with a composite of all advanced countries, rather than just the US.

This paper contributes to the literature on decoupling that focuses on the changes in the pattern of co-movements between industrial and developing countries. It complements the multicountry empirical research in the field by studying the case of India in detail.

The remainder of this paper is divided into the following sections. Section 2 discusses what economic theory and existing evidence tells us about business cycle synchronisation for developing countries. Section 3 deals with methodological issues such as business cycle identification and detrending, and also discusses our dataset and its limitations. Section 4 presents preliminary findings based on graphical analysis. Section 5 first describes the Harding-Pagan index of concordance and then discusses our main results. Section 6 presents sensitivity analyses of the results. Section 7 concludes and suggests areas for further research.

2 Business Cycle Synchronisation

There is no consensus in the theoretical literature on the impact of increasing trade and financial liberalisation on business cycle integration. Some theoretical arguments predict decoupling while others predict synchronisation. An empirical literature has sprung up, aiming to resolve this debate.

2.1 Theory

There are many channels through which synchronisation might come about. The first is the demand channel, which emphasises that demand shocks in one economy lead to income shocks in its trading partners. Thus, as intra-industry trade grows, output correlations increase leading to business cycle convergence (Frankel and Rose 1998).

The second argument emphasises financial market linkages and “contagion”. As financial integration increases, capital flows in different countries are synchronised through various channels of financial contagion including herd behaviour and information asymmetry. Region-based investment decisions and positively correlated capital shocks also lead to synchronisation (Kose et al 2003).

The third channel through which co-movement comes about between two countries is monetary policy. Significant de facto openness on the capital account is now found in almost all large countries. Under these conditions, when a country chooses to engage in exchange rate pegging, whether de facto or de jure, it loses autonomy of monetary policy. As an example, countries in west Asia have adopted US monetary policy through the use of currency boards. This induces co-movement. In the Indian case, there is evidence of periods of de facto exchange rate pegging to the US dollar (Patnaik 2007), and of the consequent loss of monetary policy autonomy (Patnaik 2005).

Commodity price movement, such as the price of oil, and remittances from industrial countries to developing countries constitute other channels through which business cycles are transmitted. However, Krugman (1993) argues that inter-industry international trade leads to specialisation. As specialisation increases, business cycles diverge due to sector-specific shocks. Similarly, increased financial integration also promotes product specialisation, as firms use portfolio diversification to insure against countryspecific shocks. This can increase business cycle asymmetry.

2.2 Empirical Evidence

The early literature focusing on developed countries found significant evidence that increasing trade integration led to increased business cycle synchronisation. Frankel and Rose (1998) estimated an instrumental variable regression model to test if bilateral trade intensity explains cyclical output correlations in the industrial world. Shin and Wang (2003) test a similar model, also controlling for intra-industry trade. Both studies show that increasing trade intensity led to increased business cycle synchronisation.

The literature on emerging markets has mixed results. Agenor et al (2000) and Rana (2008) present stylised facts to show that output correlations with developed countries have increased over time. Calderon et al (2007) present similar results, but find that controlling for production structure asymmetries between

october 31, 2009 vol xliv no 44


countries yields lower output correlations. Chan and Khong (2007) find that Asia-Pacific economies tend to be more correlated with Japan than the US, and this synchronisation between Asia-Pacific economies is also confirmed by Kumakura (2006) and Moneta and Ruffer (2009).

Some studies find evidence of decoupling. Kose et al (2003) find that increased trade and financial liberalisation adds to contagion of macroeconomic and trade shocks but the effect for developing countries is weak. Fidrmuc et al (2008) conduct cross-spectral analysis between quarterly GDP of the Organisation for Economic Cooperation and Development (OECD) countries and emerging markets such as India and China. They estimate dynamic correlations and find that over the sample period 1996-2006 there is little coherence in business cycle frequencies of India and China with the OECD. Kose et al (2008) find that while there is no strong evidence in support of worldwide convergence of business cycles, there is evidence of inter-group convergence within industrial countries and within emerging economies. This seems to suggest decoupling between industrial countries and emerging economies.

2.3 Business Cycles in India

The existing empirical literature in India in the field of business cycle analysis deals with the problems of dating the cycle, and examining leading, coincident and lagging indicators (RBI 2006; Patnaik and Sharma 2002; Dua and Banerji 2006; Chitre 2001). These studies find evidence of market-oriented cycles post-1991 and also that some indicators of world output are relevant as leading indicators of Indian cycles (RBI 2006; Mall 1999).2 Some of the studies on international business cycle synchronisation include India as one of many countries in a multi-country dataset (Kose and Yi 2001; Agenor et al 2000; Calderon et al 2007). This limits their ability to obtain greater detail on India. However, studies like Fidrmuc et al (2008) and Akin and Kose (2008) which are closer to studying business cycle synchronisation of India and China as a group, or as part of the smaller group, with the industrial world find some evidence in favour of decoupling.

3 Data Set and Definitions
3.1 Identifying the Business Cycle

We follow the National Bureau of Economic Research (NBER) approach and study the trend-cyclical component of seasonally adjusted data. However, in order to address the “classical expansion” faced by emerging markets, where all measures of output have been on a steady increase over the past decade or so, we modify this approach to study cyclical fluctuations in annualised point-on-point growth rates of output. Effectively, we are studying growth rate cycles.

This approach is based on the premise that shocks to both the trend component and the cyclical component of output are relevant to business cycle analysis. It has the advantages of not modifying data properties via detrending, and lowering the impact of possible structural breaks on the results. This is especially relevant to emerging economies, where recent work on trend cycle integration suggests a stochastic data generating process for the trend component of output (Aguiar and Gopinath 2007).

Economic & Political Weekly

october 31, 2009 vol xliv no 44

An alternative method to identify the business cycle component of an output series is to detrend it using a time or frequency domain filter.3 However, detrending can induce spurious cycles and the results are vulnerable to filter-sensitivity (Canova 1998; Harvey and Jaeger 1993). Over the past two decades, India has seen several economic and institutional changes, including in its exchange rate regime, monetary policy framework, financial regulatory framework and trade policy structure. India has globalised rapidly and witnessed rapid economic growth. Given this institutional environment, the case for trend-cycle interaction is strengthened.

3.2 Data

The literature on business cycles in India uses monthly data for industrial production as a proxy for output, for two reasons. First, structural changes in the Indian economy over the last two decades have caused monsoon-related cycles in the period 1950-91 to morph into growth/growth rate cycles in the 1990s (Patnaik and Sharma 2002). This makes studying investment-inventory cycles relevant only after 1991. Second, any meaningful analysis of cyclical fluctuations require data of quarterly or monthly frequency. This is not easily available in India. Since quarterly GDP data is available only from 1996, the use of either annual or quarterly GDP data is inadequate. Data for employment, retail sales and income are not available on a monthly or even quarterly basis. The dataset that we create runs from August 1992 till December 2008. Monthly data for the Indian index of industrial production (IIP) is obtained from the Business Beacon database published by the Centre for Monitoring Indian Economy (CMIE). We source data on merchandise exports, GDP, gross flows on the current and capital account, corporate profits after tax, and corporate revenue growth from the same database.

We use the Conference Board coincident indicator for the US. It is a composite of the IIP, non-farm payroll employment, personal disposable income excluding transfers and retail manufacturing and sales.4 We source the US IIP from the web site of the Federal Reserve.

The Advanced Economies Index of Industrial Production (Adv Ec IIP) is a weighted index of non-seasonally adjusted industrial production for 22 countries classified as “industrial” by the International Monetary Fund. The value added in industry in the year 2000 (expressed in US$) is used as a weighting factor for each country.5 This data is sourced from the IMF-IFS. Similarly, the data for world trade used in the sensitivity analysis is obtained from the IMF-IFS.

4 Exploratory Analysis

In this section, we present preliminary evidence of India’s trade and financial integration with the world economy and examine how Indian macroeconomic variables behaved during world expansions and recessions. We examine evidence of change in business cycle synchronisation across the period 1992-2008.

Industrial production indices are measures of quantity and thus represent real variables. We seasonally adjust the data using X-12 ARIMA.6 Following Frankel and Rose (1998), who break their sample into four equal parts to examine the increase in inte gration,

corporate profits all fell to very low levels.

140 16












4 1995 1990 1995 2000 2005 1995 1990 1995

the sample period is cut across into three roughly equal sub-samples. The break-points chosen are August 1997 and August 2003.

4.1 Increased Integration

There has been a sharp increase in India’s integration with the world economy on both trade and financial flows, as shown in Figure 1, which shows graphs for the growth of gross flows on the BOP to GDP, and the exports of goods and services to GDP.

Table 1 shows averages of these values for the three periods of interest.

In the context that there is no consensus in the literature on the impact of increasing trade and financial liberalisation on business cycle integration, establishing or rejecting the synchronisation hypo thesis is an important element in the

Table 1: Ratios of Trade and Gross BOP Flows to GDP in India (%)

Sub-Sample Trade to GDP Gross BOP Flows to GDP

1992-97 20.44 45.83

1997-2003 23.28 53.77

2003-08 34.26 93.94

Table 2: Correlations of Weekly Returns on the CMIE Cospi Stock Market Index against Global Stock Market Indexes

UK FTSE-100 Japan Nikkei-225 US S&P 500

1992-97 -0.008 -0.038 -0.023

1997-2003 0.184 0.168 0.167

2003-08 0.463 0.39 0.339

Full period 0.192 0.149 0.15

debate. The sharp increase in economic integration suggests that business cycle synchronisation could have changed over these periods, thus necessitating separate measurement of business cycle synchronisation.

Table 2 shows correlations of the CMIE Cospi stock market index, which depicts the total returns on the broad market in India, against three major international indexes: the US S&P 500 index, the Japanese Nikkei 225 index and the UK FTSE-100 index. With all these three indexes, across the three sub-

Figure 2: What Happened in India in the US Recession of 2001?

periods, correlations have gone up. This sug-YOY growth in PAT of companies YOY growth in India’s merchandise exports gests increasing synchronisation with the 25 60


world economy. In the latest period, the cor- 20

40relation against the UK FTSE-100 (0.463) and 15

30 the Japanese Nikkei 225 (0.39) exceeds the 10



correlation with the US S&P 500 index. 5 0 0

1/1/2000 1/2001 1/2002 1/2003 1/2004
/ 3 3/2001 3/2002 3/2003 3/2004 3/2005


4.2 Preliminary Evidence

-5 -20

To look at some preliminary evidence about YOY IIP growth YOY growth in sales of companies 11


whether business cycles in India have been


“coupled” or “decoupled” with those in indus-9


trial countries, we look back towards the last US 7 10

business cycle as defined by the NBER (starting

5 5

in March 2001 and ending in November 2001).

3 0

Figure 2 shows data for India during that pe


-5riod. This shows that the growth of exports, 1/2000 1/2001 1/2002 1/2003 1/2004 3/2000 3/2001 3/2002 3/2003 3/2004 3/2005

Since the above analysis is limited to one business cycle downturn in the US, it only presents anecdotal evidence of greater synchronisation. However, graphs for a longer sample period (Figure 3, p 113) also suggest similar behaviour. Industrial production in

2000 2005 India across business cycle peaks and troughs

over the period 1995-2008 shows increased integration. Point-on-point growth rates between the US coincident indicator and Indian IIP, as well as those between industrial production in advanced economies and in India suggest the same, especially in the sample period 2003-08.

Finally, we present two sets of correlations between Indian IIP and the two variables representing world output. The first is cross-correlations (Table 3, p 113), which are the simplest and most commonly used method to analyse co-movements between series. The second is rolling correlations across an eight-year window (see Figure 3) with Indian IIP and the US coincident indicator and Adv Ec IIP.

Despite their static nature, cross-correlations provide two sources of insight into co-movements. The level of the correlation is indicated by the strength of co-movements. The nature of pro/ counter cyclicality of the variable is indicated by the sign – a positive sign indicates pro-cyclicality while a negative sign indicates counter-cyclicality. A value of zero indicates that the variable is acyclical. It can be seen that as we move across samples, the correlations switch signs from negative to positive. They also increase considerably in magnitude and statistical significance, with all correlations in Sample 3 (2003-08) being significant at 1%.

Similarly, the rolling correlations can also be seen to be increasing with time, starting from a negative value in the mid1990s to above 0.5 post-2005 (Figure 3). This is the case for both the US coincident indicator and the Adv Ec IIP.

Overall, the preliminary analysis suggests that business cycles in the rest of the world show co-movement with Indian IIP and that the correlation between growth rates of IIP in India and the industrial economies (particularly the US) has been increasing

october 31, 2009 vol xliv no 44

Figure 3: Preliminary Evidence

YOY IIP growth 11


0.60000 9

US Coincident









AEIIP 0.00000

6/1995 6/1996 6/1997 6/1998 6/1999 6/2000 6/2001 6/2002 1

-0.10000 1/2000 1/2001 1/2002 1/2003 1/2004

0.03000 0

0.00500 0




US Coincident 0.00400

0.00600 0.02000 0


0.02000 0.00200


0.01000 0 0.00100





0.00000 0.00000 0




8/1992 8/1994 8/1996 8/1998 8/2000 8/2002 8/2004 8/2006 8/1992 8/1994 8/1996 8/1998 8/2000 8/2002 8/2004 8/2006 -0.00200

-0.01000 0 -0.00200




INIIP -0.00300



-0.02000 0





-0.03000 0 -0.00500




over time. Section 5 now turns to a more formal analysis of this However, given the markov-transition probability structure of preliminary finding. recessions (Pr(St+1=0, S=0)) >> (Pr(St+1=0, S=1)), there is ex

t t

tensive serial correlation in the S series (Harding and Pagan


5 Empirical Analysis 2006). Also, since the data duration is very short, the chances of a prolonged expansion or recession in one of the series skewing the

5.1 Methodology

value of the index are non-zero. There are a variety of formal methods in the literature to study To correct for these flaws, Harding and Pagan (2006) demonstrate business cycle synchronisation, the most popular being dynamic that the following relationship holds between the correlation cocorrelations, spectral analysis and Harding-Pagan’s index of concor-efficient ρˆ between S and S and ˆI which implies that the


dance (Simone 2003; Fidrmuc et al 2008; Chan and Khong 2007). properties of ρˆ are symmetric to that of Ì

xy xy

Table 3: Cross-correlations with Indian IIP

ˆIxy = 1+ 2ρˆxyσˆSxσˆSy + 2μˆSxˆμSy – μˆSx – μˆSy

Variables t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4

Period 1: 1992-97

To estimate the correlation coefficient

Coincident -0.15 -0.08 -0.04 0.09 -0.07

US -0.02 -0.07 0.02 0.00

ρˆ , we use the following OLS estimation:

Adv Ec IIP 0.15 0.13 -0.01 -0.01 0.01 -0.06 -0.13 -0.30* -0.26 xy

Period 2: 1997-2003

S/(σˆ)(σˆ) = A + (ρ) (S)/ (σˆ)


US Coincident 0.07 0.14 0.2 0.25* 0.26* 0.30** 0.29** 0.29** 0.30**

(σˆ) +


Adv Ec IIP 0.16 0.27* 0.35*** 0.46*** 0.48*** 0.46*** 0.52*** 0.49*** 0.47***

Period 3: 2003-08

where S denotes the sample standard

US Coincident 0.37*** 0.47*** 0.44*** 0.54*** 0.56*** 0.56*** 0.44*** 0.38*** 0.35***yt

deviation of S. Given that ε inherits the

Adv Ec IIP 0.36*** 0.31** 0.50*** 0.45*** 0.55*** 0.41*** 0.45*** 0.36*** 0.33*** yttFull period: 1992-2008 serial correlation in S, we report p-values


US Coincident 0.14 0.15* 0.17* 0.19** 0.20** 0.22** 0.18** 0.19** 0.15*

for the Heteroskedasticity-Autocorrela

Adv Ec IIP 0.20** 0.20** 0.21** 0.21** 0.24*** 0.19** 0.18** 0.1 0.1

tion (HAC) corrected t-statistics for ρˆ .7


We use the index of concordance (HP Index) as developed by Harding and Pagan (2006) as a means to test increasing busi-5.2 Main Results ness cycle synchronisation across our three sample periods. This The results of the Harding-Pagan analysis on the data and three measures the proportion of the time that two variables are in the sub-samples are reported in Table 4 (p 114). We report the index of same state. Assuming two variables x and y over N time periods, concordance and the cross-correlations of the state variables as two the index of concordance between them would be: measures of concordance. Newey West HAC t-statistics and p-values Ì = #[S=1, S=1] + #[S=0, S=0]reported are for the estimated correlation coefficient, but can be


extrapolated to the index of concordance (Harding and Pagan

N The value of the HP index ranges between 0 and 1. An index 2006). The results support the early exploratory analysis: there is value of close to 1 would indicate perfect procyclicality while an business cycle synchronisation between India and the rest of the index value of 0 would indicate perfect counter-cyclicality. world, and that synchronisation has been increasing with time.

For the full sample (1992-2008) the index of concordance sug- Asia liberali sed trade and financial policy. However, they
gests that Indian and US business cycles are in the same phase for contrast sharply with Fidrmuc et al (2008) who find evidence of
63.9% of the sample period, while cycles of industrial production Chinese and Indian decoupling from the OECD countries using
in India and advanced economies are in the same phase for spectral analysis.
74.3% of the sample. Both are statistically significant at a 95% In the following sections, we test the sensitivity of these results
Table 4: Harding-Pagan Index of Concordance with Indian IIP through a series of alternative estimation procedures.
Variable I xy S x S y t-statistic p-value Period 1: 1992-97 6 Sensitivity Tests
US Coincident Ind 0.536 -0.136 -0.8 0.427 We present the robustness of our main results to four sets of sen-
Adv Ec IIP 0.5 -0.333 -2.629 0.011** sitivity tests:
Period 2: 1997-2003 US Coincident Ind 0.767 0.356 1.544 0.127 (1) The first is the redefinition of sample periods. While we show
Adv Ec IIP 0.781 0.526 2.72 0.008** evidence of synchronisation across time, we believe that there is
Period 3: 2003-08 no clear “begin” or “end” date for this synchronisation, rather
US Coincident Ind 0.781 0.501 6.438 0.000*** that it is a slowly evolving phenomenon that reflects changes in
Adv Ec IIP 0.984 0.965 43.497 0.000*** the underlying structural composition of the Indian economy
Full period: 1992-2008 with respect to the rest of the world. For the analysis we change
US Coincident Ind 0.639 0.254 2.178 0.031** Adv Ec IIP 0.743 0.476 3.569 0.000*** the sub-sample break points to February 1998 and June 2004. (2) The second sensitivity test is done by changing the method
Table 5: Sensitivity Analysis 1: Harding Pagan Analysis with Changed Sub-samples used for analysis by detrending the data, rather than conducting
Variable I xy S x S y t-statistic p-value growth rate cycle analysis. We have so far conducted all analysis
Period 1: 1992-97 on the trend-cyclical component of output. We now detrend the
US Coincident Ind 0.597 -0.075 -0.451 0.654 data using the Hodrick-Prescott (HP) filter, which is widely used
Adv Ec IIP 0.565 -0.277 -2.444 0.017** Period 2: 1997-2003 in business cycle literature.
US Coincident Ind 0.636 0.196 0.941 0.35 (3) The third test is to utilise another methodology that is widely
Adv Ec IIP 0.779 0.534 3.136 0.002*** used for measuring co-movement: spectral analysis (Fidrmuc et
Period 3: 2003-08 al 2008; Calderon et al 2007).
US Coincident Ind 0.453 0.277 2.325 0.024** (4) Finally, we verify that the main results hold across redefinini-
Adv Ec IIP 0.396 0.212 2.244 0.029** Full period: 1992-08 tion of some key variables.
US Coincident Ind 0.639 0.254 2.178 0.031** Adv Ec IIP 0.743 0.476 3.569 0.000*** 6.1 Redefining Sample Periods
Table 5 presents the results of the Harding-Pagan analysis for the
confidence interval, and the value for the US is lower. This indi changed sample periods. The key results hold. One difference is
cates business cycle synchronisation. the value of the index of concordance for Adv Ec IIP in Sample 3
The most recent sample (2003-08) shows stronger synchroni – it seems to have fallen considerably although it remains statisti
sation. The index rose to 0.781 with the US coincident indicator, cally significant at 5%.
and as high as 0.984 against advanced countries.
In Period 1 (1992-97), both the US coincident indicator and 6.2 Detrending
IIP for advanced economies were negatively correlated with The Hodrick-Prescott filter is a time-domain filter that renders
Indian industrial production, suggesting that the Indian business the resulting cyclical component stationary.8 We use the Hodrick
cycle was weakly counter-cyclical to the world during this time. Prescott filter with a smoothing parameter of 14,400 since the
However, this was a high volatility period due to structural ad data is of a monthly frequency in order to decompose the series
justment to reforms and revival from the balance-of-payments into trend and cycle. Our empirical strategy is then repeated
crisis of 1991. Hence, it can be viewed as a “transition period” in using the detrended data.
the Indian economy, a possible explanatory factor for this result. Table 6 (p 115) reports these results. While these results cannot
Also, across all samples, it can be seen that the Adv Ec IIP is be directly contrasted with our main findings (this analysis tests
more strongly correlated with Indian IIP, suggesting that the for growth cycle synchronisation, while our main results test for
Indian synchronisation with industrial economies as a whole is growth rate cycle synchronisation), they still examine broadly the
stronger than the synchronisation with the US. In fact, for the last same question of synchronisation in the context of integration.9
period 2003-08, the index of concordance against Adv Ec IIP is as We see that the synchronisation of business cycles in the most
high as 0.984, and it has a t statistic of 43.5. recent sample (2003-08) is robust to the HP filter. However, there
Our results support Calderon et al (2007), who test for the are two notable differences in the results obtained. First, the
impact of increasing trade intensity on business cycle synchroni world variable Adv Ec IIP is not significantly synchronised with
sation and find increased correlations for countries that have Indian IIP across the total sample 1992-2008. Second, the HP filter
closer trade ties. They are also similar to those of Rana (2008) finds that there is no statistically significant synchronisation in
who also finds increased synchronisation between east Asian the period 1992-97. This agrees with evidence of negative syn
economies and the rest of the world in the time period that east chronisation in this period.
114 october 31, 2009 vol xliv no 44 Economic & Political Weekly


Table 6: Sensitivity Analysis 2: Harding-Pagan Analysis with HP-Filtered IIP Series Table 7: Sensitivity Analysis 3: Harding-Pagan Analysis (Redefining Key Variables)
Variable I xy S x S y t-statistic p-value Variable I xy S x S y t-statistic p-value
Period 1: 1992-97 Period 1: 1992-97
US Coincident Ind 0.41 0.116 0.843 0.402 US IIP 0.375 -0.064 -0.266 0.791
Adv Ec IIP 0.328 -0.186 -0.766 0.447 World trade 0.536 -0.299 -2.57 0.013**
Period 2: 1997-2003 Period 2: 1997-2003
US Coincident Ind 0.904 0.758 12.618 0.000*** US IIP 0.781 0.548 2.966 0.004***
Adv Ec IIP 0.507 0.218 1.762 0.082* World trade 0.795 0.509 2.893 0.005***
Period 3: 2003-08 Period 3: 2003-08
US Coincident Ind 0.954 0.776 6.445 0.000*** US IIP 0.429 0.123 0.913 0.365
Adv Ec IIP 0.862 0.243 1.775 0.081* World trade 0.841 0.698 7.282 0.000***
Full period: 1992-2008 Full period: 1992-2008
US Coincident Ind 0.629 0.242 1.672 0.096* US IIP 0.516 0.048 0.353 0.724
Adv Ec IIP 0.599 0.238 1.595 0.112 World trade 0.705 0.397 3.014 0.003***
6.3 Dynamic Correlations period 1992-2008, i e, the period that saw a significant rise in
We now use spectral analysis as a methodology for checking co India’s trade and capital flows. Finally, the linkages of the Indian
movement. It is widely used in the literature (Fidrmuc et al 2008; economy are stronger when measured against a broad set of
Calderon et al 2007) and the results confirm our key findings. industrial countries as opposed to just the US.
Spectral coherence (also called dynamic correlation) for each world This paper contributes to the evolving empirical evidence on
variable with respect to Indian IIP has substantially increased over the the question of whether emerging market economies such as
period 1992-2008, over growth rate cycle frequencies. The Appendix India are decoupled with industrial economies or not. As there is
reports our findings and methodology in greater detail. no consensus in the literature, and business cycles in India have
emerged as an important part of the debate, the paper is an
6.4 Redefining Key Variables important contribution as it strongly supports the evidence that
For the final sensitivity test, we redefine our measure of US business cycles in India are coupled with those in industrial coun
business cycles from the US coincident indicator to the IIP in the tries and that this coupling has been increasing with India’s
US. In a similar vein, we use a measure of total world trade greater globalisation.
(exports plus imports), as a proxy for Adv Ec IIP.10 The results are This paper focused on establishing business cycle synchronisa
reported in Table 7. tion. It did not attempt to study the transmission mechanism and
It can be seen that all results hold with respect to the variable causal relationships through which business cycle synchronisation
measuring world trade, but there is a fall in the statistical signi takes place. In a similar vein, it did not control for intervening
ficance of US industrial production both in the total sample and in countries, such as other emerging markets that may help precipitate
the third period (2003-08). or dissipate the impacts of shocks, thereby strengthening or weak
ening co-movements between India and the advanced economies.
7 Conclusions Finally, it analysed only output fluctuations between two coun-
In this paper, we find that the Indian business cycle is synchro tries to study co-movement of cycles. Future research would need
nised with that of the US and other industrial economies. We also to analyse other countries and other variables as well as study
find that this synchronisation has increased across time in the the transmission mechanism of co-movements.


1 Exports to the US accounted for 13% of India’s exports in 2007, and it has long been India’s biggest trade partner.

2 Most of these studies look at growth cycles, i e, deviations of output from a designated “trend growth”.

3 Commonly used filters include the Hodrick-Prescott filter, the Baxter-King filter and the Christiano-Fitzgerald filter.

4 This indicator is available from the Conference Board’s web site at http://www.conference-board. org/economics/bci/

5 Bases are harmonised to 2000=100 using chainlinking via ratio-splicing. 6 Model specifications were verified using the HEGY seasonal unit root tests and residual diagnostics.

7 To identify turning points, we use the Harding-Pagan turning points algorithm as implemented in the software GROCER (Dubois and Michaux 2008).

8 Criticisms of the HP filter include spurious cycles, phase shifts in the variables and a high level of sensitivity of results (Canova 1998; Harvey and Jaeger 1993).

9 See Harding and Pagan (2002) for an overview of the differences between growth and growth rate cycles.

10 Export and import data is sourced from the IMF-IFS, and expressed in $ billion. US IIP is sourced from the St Louis Federal Reserve Database. Both variables are adjusted for seasonal fluctuations.


Agenor, Pierre-Richard, C John McDermott and Eswar S Prasad (2000): “Macroeconomic Fluctuations in Developing Countries: Some Stylised Facts”, World Bank Economic Review, 14(2):251-85.

Aguiar, Mark and Gita Gopinath (2007): “Emerging Market Business Cycles: The Cycle Is the Trend”, Journal of Political Economy, 115: 69-102.

Akin, C and M A Kose (2008): “Changing Nature of North-South Linkages: Stylised Facts and Explanations”, Journal of Asian Economics, 19(1): 1-28.

Baxter, M and M A Kouparitsas (2005): “Determinants of Business Cycle Co-movement: A Robust Analysis”, Journal of Monetary Economics, 52(1): 113-57.

Calderon, Cesar, Alberto Chong and Ernesto Stein (2007): “Trade Intensity and Business Cycle Synchronisation: Are Developing Countries Any Different?” Journal of International Economics, 71(1): 2-21.

Canova, Fabio (1998): “Detrending and Business Cycle Facts”, Journal of Monetary Economics, 41(3): 475-512.

Chan, Tze-Haw and Wye Leong Roy Khong (2007): “Business Cycle Correlation and Output Linkages among the Asia-Pacific Economies”, MPRA Paper 11305 (Germany: University Library of Munich).

Chitre, Vikas (2001): “Indicators of Business Recessions and Revivals in India: 1951-1982”, Indian Economic Review, 36(1): 79-105.

Darvas, Zsolt and Gyrgy Szapry (2008): “Business Cycle Synchronisation in the Enlarged EU”, Open Economies Review, 19(1): 1-19.

Dua, Pami and Anirvan Banerji (2006): “Business Cycles in India”, Working papers 146, Centre for Development Economics, Delhi School of Economics.

Dubois, Eric and Emmanuel Michaux (2008): Grocer

1.3: An Econometrics Toolbox for Scilab. Fidrmuc, Jarko, Iikka Korhonen and Ivana Btorov (2008): “China in the World Economy: Dynamic Correlation Analysis of Business Cycles”, BOFIT Discussion Papers 7/2008, Bank of Finland, Institute for Economies in Transition. Frankel, Jeffrey A and Andrew K Rose (1998): “The Endogeneity of the Optimum Currency Area Criteria”, Economic Journal, 108(449): 1009-25. Harding, Don and Adrian Pagan (2002): “Dissecting

the Cycle: A Methodological Investigation”, Journal of Monetary Economics, 49(2): 365-81.

october 31, 2009 vol xliv no 44

– (2006): “Synchronisation of Cycles”, Journal of Kose, M Ayhan, C Otrok, E Prasad (2008): “Global – (2007): “India’s Currency Regime and Its Conse-Econometrics, 132(1): 59-79. Business Cycles: Convergence or Decoupling?” quences”, Economic & Political Weekly.

Harvey, A C and A Jaeger (1993): “Detrending, Stylised (Massachusetts, USA: National Bureau of Eco-Patnaik, Ila and Rachna Sharma (2002): “Business Facts and the Business Cycle”, Journal of Applied nomic Research Cambridge). Cycles in the Indian Economy”, Margin, 35(1): 71-79. Econometrics, 8(3): 231-47. Krugman, Paul (1993): The Transition to Economic and Prasad, Eswar S (1999): “International Trade and the

Iacobucci, Alessandra (2003):. “Spectral Analysis for Monetary Union in Europe, Cambridge University Business Cycle”, Economic Journal, 109(458): 588-606.

Economic Time Series: Documents de Travail Press, pp 241-61. Rana, Pradumna B (2008): “Trade Intensity and

de l’OFCE 2003-07”, Observatoire Francais des Kumakura, Masanaga (2006): “Trade and Business Business Cycle Synchronisation: The Case of East

Conjonctures Economiques. Cycle Co-movements in Asia-Pacific”, Journal of

Asian Countries”, Singapore Economic Review

Kohn, D L (2008): “Global Economic Integration and Asian Economics, 17(4): 622-45.

(SER), 53(02): 279-92.

Decoupling”, Speech at the International Research Mall, O P (1999): “Composite Index of Leading Indica-RBI (2006): “Leading Indicators for the Indian Econo-Forum on Monetary Policy, Frankfurt, Germany. tors for Business Cycles in India”, Technical Remy”, Technical Report (Mumbai: Reserve Bank of Kose, M Ayhan and Kei-Mu Yi (2001): “International port 3, Reserve Bank of India.

India). Trade and Business Cycles: Is Vertical Specialisation Moneta, Fabio and Rasmus Ruffer (2009): “Business

Shin, Kwanho and Yunjong Wang (2003): “Trade Intethe Missing Link?”, American Economic Review, Cycle Synchronisation in East Asia”, Journal of gration and Business Cycle Synchronisation in 91(2): 371-75. Asian Economics, 20(1): 1-12. East Asia”, Asian Economic Papers, 2(3): 1-20. Kose, M Ayhan, Eswar S Prasad and Marco E Terrones Patnaik, Ila (2005): “India’s Experience with a Pegged Simone, Francisco Nadal-De (2003): “Common and (2003): “How Does Globalisation Affect the Exchange Rate” in Suman Bery, Barry Bosworth and Idiosyncratic Components in Real Output: Fur-Synchronisation of Business Cycles?”, American Arvind Panagariya (ed.), The India Policy Forum 2004, ther International Evidence”, IMF Working Papers Economic Review, 93 (2): 57-62. Brookings Institution Press and NCAER, pp 189-226. 02/229, International Monetary Fund.

Figure A.1: Cross-Spectral Analysis between Indian IIP and World Variables co-movements into those at short, me-

Smoothed periodogram of INIIP and USCOINCIDENT (log (Spectrum)) Smoothed periodogram of INIIP and AEIIP (log (Spectrum))

dium and long-term frequencies. How


-8 ever, as we are studying growth rate -13

-10 cycles, we must keep in mind that differencing is an asymmetric frequency

-15 -12

operation and may lead to the intro

-17 -14

duction of high frequency compo

-19 -16

nents (Iacobucci 2003). Moreover, since our data duration is short (197


-18 0.0625 0.5625 1.0625 1.5625 2.0625 2.5625 3.0625 3.5625 4.0625 4.5625 5.0625 5.5625 0.06 0.69 1.31 1.94 2.56 3.19 3.81 4.44 5.06 5.69 observations across 16 years of

Frequency bandwidth = 0.158 Frequency bandwidth = 0.158 monthly data), spectral estimations Squared coherency Squared coherency

may be biased. In spite of these limi


0.06 0.56 1.06 1.56 2.06 2.56 3.06 3.56 4.06 4.56 5.06 5.56 0.0625 0.5625 1.0625 1.5625 2.0625 2.5625 3.0625 3.5625 4.0625 4.5625 5.0625 5.5625

modified Daniell filter to generate the

Frequency Frequency

periodogram with scaled densities.


Second, we report spectral coherence (Table A.1), a frequency

1 Data Sources

domain analogue to the correlation coefficient. It is calculated as per

Table 8: Variable Definitions

Equation 4 where S1(k) is

Table A.1: Spectral Coherence with Indian IIP

Variable Source Unit Frequency Time period

the spectral periodogram

Variable Spectral Coherence

Indian IIP CMIE BB Index (1993=100) Monthly 1992-2008

of Variable 1 at frequency Low Freq Mid Freq High Freq

Gross BOP flows CMIE BB Percentage Quarterly 1985-2008 k, S2(k) is that of vari-Period 1: 1992-97

to GDP: India growth

able 2 and S12(k) is their US Coincident Ind 0.07 0.04 0.04 Indian exports to GDP CMIE BB Percentage growth Quarterly 1985-2008

cross-spectrum. Based on Adv Ec IIP 0.03 0.02 0.06 CMIE COSPI Index CMIE BB Index Weekly 1992-2008 growth rate cycle perio-

Period 2: 1997-2003 Indian corporate profit CMIE Prowess Percentage Quarterly 2000-2005 dicities (roughly between US Coincident Ind 0.08 0.18 0.15

after tax growth

12 and 24 months), low

Adv Ec IIP 0.05 0.64 0.35

Indian merchandise CMIE BB Percentage Quarterly 2000-2005 frequencies are identi-Period 3: 2003-08

exports growth

fied as those with 0.5 or

US Coincident Ind 0.1 0.69 0.51

Indian corporate CMIE Prowess Percentage Quarterly 2000-2005 less cycles per year, mid-Adv Ec IIP 0.26 0.68 0.42

revenue growth growth range as those between 1 Full period: 1992-2008

US coincident The Conference Index Monthly 1992-2008 and 0.5 cycles per year indicator Board (2004=100) US Coincident Ind 0.02 0.12 0.08

and high greater than 1 Advanced economies IMF-IFS Index Monthly 1992-2008 cycle per year. Adv Ec IIP 0.21 0.11 0.19 IIP (1999=100)

Kˆ (k)=S12(k)/ √ Sˆ (k) Sˆ (k)

US S& P Yahoo Finance Index Weekly 1992-2008 1212

UK FTSE-100 Yahoo Finance Index Weekly 1992-2008

It can be seen that both sets of results indicate business cycle

Japan Nikkei-225 Yahoo Finance Index Weekly 1992-2008

synchronisation, and the mean coherence estimates over the three US IIP FRED Database Index (1999=100) Monthly 1992-2008 sub-samples indicate that this synchronisation has been increasing World trade IMF-IFS USD Billion Monthly 1992-2008 over time. Following RBI (2006), we consider a coherence of greater

that 0.30 as an indication of significant co-movement. It can be seen

2 Spectral Analysis

that the coherence at the mid-range frequencies over the period Spectral analysis provides a frequency domain complement to cross-2003-08 are 0.69 and 0.68, respectively, and that coherence across correlation analysis, with the advantages of being able to decompose this range of frequencies has been increasing over the period 1992-2008.

116 october 31, 2009 vol xliv no 44

tations, a frequency domain perspec-tive does provide further insight into business cycle co-movements. First, we present cross-spectral periodograms (Figure A.1). Periodograms are estimat-ed via a Discrete Fast Fourier Trans-formation, and then smoothed with a 0 0.2 0.4 0.6 0 0.2 0.4 0.6 0.8

Dear Reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Back to Top