1.5 The Group deliberated on several issues regarding the development of the leading economic indicators during the two meetings held in Mumbai. In the First Meeting, graced by Deputy Governor Dr. Rakesh Mohan, the issues regarding the selection of reference series and methodology for construction of the composite leading economic indicators were discussed. In the Second Meeting chaired by Executive Director Dr. R. B. Barman, the empirical results, based on the suggestions received in the First Meeting were discussed. Based on these deliberations and the background empirical work undertaken, the Group finalised the Report. The Report is organized as follows. The Report is divided into six sections. Section II provides a brief description of Business Cycles. Section III covers a brief discussion about the methodologies. Section IV discusses the application of leading economic indicators, both, international and Indian and experiences gained. The details of empirical work done for estimation of cyclical components and construction of the Composite Index of Leading Indicators using growth cycle and growth rate cycle methodologies are discussed in Section V. In Section VI, Composite Index of Coincident Indicators is prescribed. Section VII concludes with Recommendations of the Group.

I.3 Acknowledgements

Table 11: Prediction evaluation of the IIP- Growth Rate Cycle

5.32. The movement of the growth cycles in the quarterly non-agricultural GDP in India is presented in Chart 8. For convenient, the data presented in the Chart, covers the period 1992-93:Q3 to 2005-06:Q4. The growth cycles for the quarterly reference series are estimated based on the deseasonalized, detrended and smoothened series. The growth cycles have been estimated using the HP- filter with parameter λ = 1600.

Chart 8: Movement of Quarterly Non-Agriculture GDP Growth Cycles

The dates of business cycles on the quarterly non-agricultural GDP series are determined using the Bry- Boschan rules and are presented in Table 12. During the reference period, 3 growth cycles have been identified. The average duration of the recessions phase is found to be 5 quarters, while the expansions phases are of found to be slightly lower at 4 quarters on average basis. The average duration of a complete cycle (trough-to-trough) is 10 quarters.

Table 12: Growth Cycle Chronology for Non-Agriculture GDP (Period in quarters)

5.34. Table 13 presents the set of initially selected twelve potential leading economic indicators based on the cross-correlations analysis²¹. From the table, Currency with the Public is found to be the most potential leading indicator, having a correlation coefficient of 0.66 with NAGDP with a lead period of 2-quarters. Similarly the correlation coefficient between Cargo handled in the major ports with a lead period of 1-quarter with NAGDP is found to be 0.66. But from policy point of view, it is necessary to have some idea about the future movement of the economy at least 2- quarters ahead. It can be observed from the cross- correlation table that, even at a lead period of 2-quarters, the correlation coefficient of Cargo handled in the major ports, IIP- electricity and Non-oil Imports with NAGDP are found to be 0.65, 0.52 and 0.56, respectively.

Table 13: Potential Leading Indicators for NAGDP – Growth Rate Cycle

5.35. From the initially selected potential leading economic indicators, attempts were made to select the optimum combination of the indicators, which can track the movement of the NAGDP at least 2- quarters ahead. The optimum CILI found to have been able to track the movement in the NAGDP consists of five indicators, viz., Currency with the Public, WPI- Manufactured Products, Production of Commercial Motor Vehicles, IIP- Electricity and Non-oil Imports. The CILI is constructed as the simple average of these five indicators.

²¹ See Annex-VII.D.

5.36. The CILI reveals that it has been able to capture the turning points of the non-agricultural GDP. Chart 9, presents the movement of the CILI and the reference series from 1992-93:Q3 to 2005-06:Q4.

Chart 9: Movement of the CILI and Non-Agriculture GDP Growth Cycles

5.37. The performance of the CILI has been presented in Table 14. On an average, the CILI leads the non-agricultural GDP cycle peaks by 2 quarters and troughs by 5 quarters. It is being observed that the CILI has been able to track all the trough of the growth cycles at least 2-quarters in advance. However, the CILI could not capture the peak during 2000-01:Q1 in advance.

Table 14: Lead Record of the Composite Leading Index for the Indian Economy

5.38. Table 15 presents the forecasting power of the probit model in classifying the different phases correctly for different threshold probability level. Corresponding to a threshold level of 0.5, in the expansion phases, 19 out of 24 observations and, in the recession phase, 14 out of 21 observations are correctly classified by the model, i.e., the estimated model correctly forecasts 79.2% of the observations in the expansion phase and 66.7% observations in the recession phase. Overall, the model correctly forecasts 73.3% of the total number of observations. When the threshold level is considered as 0.4, the model correctly forecasts 70.8% and 76.2% of the observations correctly in the expansion and recession phase respectively.

Table 15: Prediction evaluation of the NAGDP- Growth Cycle

5.39. As an alternative, an attempt has been made to track the movement of the different phases of the economy by the application of the Principal Component Analysis. For this purposes, five economic indicators were selected, viz., CWP, WPI- Manufactured Product, CMV, IIP- Electricity and Non-oil Imports. The first three principal components could explain 99.6 per cent of the total variation present in the data. The CILI based on Principal Components (CILI_PC) is constructed as the simple average of these three principal components.

5.40. The CILI_PC, constructed based on the above-mentioned methodology, reveals that it has been able to capture the turning points of the Non-Agricultural GDP. However, the CILI_PC generated two extra turning points, viz., one as trough at 1999-00:Q4 and another as peak during 2001-02:Q2. The movements of the CILI_PC and the NAGDP are presented in Chart 10.

Chart 10: Movement of the CILI_PC and Non-Agriculture GDP Growth Cycles

5.41. The performance of the CILI_PC has been presented in Table 16. On an average, the CILI_PC leads the non-agricultural GDP cycle peaks by 3 quarters and troughs by 2 quarters. It is being observed that the CILI_PC has been able to track all the troughs of the growth cycles at least 1-quarters in advance. However, the CILI_PC could not capture the peak observed during 1998-99:Q1 in advance. The other two peaks observed during the sample period was well captured by the CILI_PC at least 3- quarters ahead.

Table 16: Lead Record of the Composite Leading Index for the Indian Economy

5.42. Table 17 presents the forecasting power of the probit model, based on CILI_PC, in classifying the different phases correctly for different threshold probability level. Corresponding to a threshold level of 0.5, in the expansion phases, 18 out of 24 observations and in the recession phase, only 7 out of 21 observations are correctly classified by the model, Thus the estimated model correctly forecasts 75.0% of the observations in the expansion phase and only 33.3% observations in the recession phase. Overall, the model correctly forecasts 55.6% of the total number of observations. When the threshold level is considered as 0.4, the model correctly forecasts 37.5% and 85.7% of the observations correctly in the expansion and recession phase respectively. Overall, the performance of the model is not found to be satisfactory.

Table 17: Prediction evaluation of the NAGDP- Growth Cycle

5.43. The movement of the growth rate cycles in the quarterly non-agricultural GDP in India is presented in Chart 11. The data presented in the Chart covers the period from 1993-94:Q3 to 2005-06:Q4. The growth rate cycles for the reference series are obtained by smoothing the annual point-to-point growth rates by a weighted moving average of period 3. The current quarter has been assigned with a weight of 3. The weight assigned for the previous quarter has been fixed at 2 and so on. The length of the moving average has been determined from the spectral density estimates of the growth rate of the reference series.

Chart 11: Movement of Quarterly Non-Agriculture GDP Growth Rate Cycles

5.44. The chronology of business cycles on the quarterly non-agricultural GDP series is presented in Table 18. During the period 1993-94:Q3 to 2005-06:Q4, 2 growth rate cycles have been identified. The average duration, for both the recessions and expansion phase are found to be 5 quarters on average basis. The average duration of a complete cycle (Peak-to-Peak) is approximately of 9 quarters.

Table 18: Growth Rate Cycle Chronology for Non-Agriculture GDP (Period in quarters)

5.45. Table 19 presents the set of nine potential leading economic indicators selected initially based on the cross-correlations analysis²². From the table, Non-food credit is found to be the most potential leading indicator, having a correlation coefficient of 0.62 with NAGDP with a lead period of 1-quarter only. Similarly the correlation coefficient between IIP- Intermediate Goods with a lead period of 1-quarter with NAGDP is found to be 0.54. But from policy point of view, it is necessary to have some idea about the future movement of the economy at least 2- quarters ahead. It can be observed from the cross- correlation table that, even at a lead period of 2-quarters, the correlation coefficient between NFC and IIP- Intermediate Goods with NAGDP are found to be 0.56 and 0.48 respectively.

Table 19: Potential Leading Indicators for NAGDP – Growth Rate Cycle

5.46. From the initially selected potential leading economic indicators, attempts were made to select the optimum combination of the indicators, which can track the movement of the NAGDP at least 2- quarters ahead. The optimum CILI, constructed as the simple average of the indicators, found to have been able to track the movement in the NAGDP consists of four indicators, viz., Non-food Credit, WPI- Food Articles, IIP- Capital Goods and US GDP.

²². See Annex-VII.D.

5.47. The CILI, constructed based on the above-mentioned methodology, reveals that it has been able to capture the turning points of the NAGDP. Chart 12, presents the movement of the CILI and the reference series from 1993-94:Q3 to 2005-06:Q4.

Chart 12: Movement of Quarterly Non-Agriculture GDP and CILI – Growth rate cycles

5.48. The performance of the CILI has been presented in Table 20. On an average, the CILI leads the non-agricultural GDP cycle peaks and troughs by 5 quarters each.

Table 20: Lead Record of the Composite Leading Index for the Indian Economy

5.49. Table 21 presents the forecasting power of the CILI, estimated through the probit model in classifying the different phases correctly for different threshold probability level. Corresponding to a threshold level of 0.5, in the expansion phases, 26 out of 29 observations and, in the recession phase, 10 out of 14 observations are correctly classified by the model, i.e., the estimated model correctly forecasts 86.2% of the observations in the expansion phase and 71.4% observations in the recession phase. Overall, the model correctly forecasts 83.7% of the total number of observations. When the threshold level is set as 0.4, the model correctly forecasts 86.2% and 78.6% of the observations correctly in the expansion and recession phase respectively.

Table 21: Prediction evaluation of the IIP- Growth Rate Cycle

5.50. As an alternative, an attempt has been made to track the movement of the different phases of the economy by the application of the Principal Component Analysis. For this purposes, five economic indicators were selected, viz., NFC, WPI- Food Articles, IIP- Capital Goods, Inter-Bank Forward Premia of US Dollar (6-months) and USGDP. The first three principal components could explain 86.1 per cent of the total variation present in the data. The CILI based on Principal Components (CILI_PC) is constructed as the simple average of these three principal components.

5.51. The CILI_PC, constructed based on the above-mentioned methodology, reveals that it has been able to capture the turning points of the Non-Agricultural GDP. The movements of the CILI_PC and the NAGDP from 1993-94:Q3 to 2005-06:Q4 are presented in Chart 13.

Chart 13: Movement of Quarterly Non-Agriculture GDP and CILI_PC (based on Principal Component)– Growth rate cycles

5.52. The performance of the CILI_PC has been presented in Table 22. It can be observed that the movement in the CILI_PC is almost same as that of the movement of the CILI, constructed based on the four indicators.

Table 22: Lead Record of the CILI (PC- based) for the Indian Economy

5.53. The forecasting power of the CILI_PC, estimated through a probit model in classifying the different phases correctly for different threshold probability level, is presented in Table 23. Corresponding to a threshold level of 0.5, in the expansion phases, 25 out of 26 observations and, in the recession phase 4 out of 8 observations are correctly classified by the model, i.e., the estimated model correctly forecasts 96.2% of the observations in the expansion phase and 50.0% observations in the recession phase. Overall, the model correctly forecasts 85.3% of the total number of observations. When the threshold level is set as 0.4, the model correctly forecasts 92.3% and 62.5% of the

Table 23: Prediction evaluation of the IIP through CILI_PC- Growth Rate Cycle

V.4.1. Monthly Index of Industrial Production – Growth Rate Cycle

5.55. For the construction of the monthly Composite Index of Coincidental Indicators (CICI), the indicators are selected based on the maximum contemporaneous correlation of the indicators with the monthly IIP series. The CICI is constructed the same way as that of the CILI, i.e., simple average of the standardized indicators. Among all the coincidental indicators, the CICI is constructed based on those indicators only, whose combination has able to track the movement of the IIP series. The indicators selected in this case are, IIP – Intermediate Goods, IIP – Consumer Durables and Sensex.

5.56. Chart 14, presents the movement of the CICI and IIP growth rate cycles. From Table 24, it can be observed that the CICI has been able to track all the trough points correctly, while the CICI missed the peak observed during September 1995. However, the movement of CICI with that of the IIP are not seems to be very regular after mid 2001.

Chart 14: Movement of the CICI and IIP Growth Rate Cycles

Table 24: Record of the Composite Coincidental Index

5.57. Based on the methodology explained in Section- III, the coincidental indicators selected for the non-agricultural GDP are found to be, IIP – General Index, Exports and Bank Credit -SCB.

5.58. Chart 15, presents the movement of the CICI and Non-Agricultural GDP growth rate cycles. From the Chart, it has been observed that the movement of the CICI moves in tandem with that of the Non-Agricultural GDP. From Table 25, it can be observed that the CICI has been able to track the turning points correctly on an average.

Chart 15: Movement of the CICI and NAGDP Growth Rate Cycles

Table 25: Record of the Composite Coincidental Index for the Indian Economy

6.1. The empirical exercise was carried out using the techniques adopted by the international organisations and also based on the methodologies suggested in some of the research papers. Broadly the methodologies could be summarised as Growth cycle methodology and Growth rate cycle methodology. Selection of the indicators was based on simple correlation.

6.2. The Composite Index of Leading Indicators (CILI) for the Index of Industrial Production (IIP) growth cycle, has been constructed based on four economic indicators, viz., Currency with the Public, WPI- Manufactured Product, IIP- electricity and Cargo handled in the major ports. The CILI is found to have a forecast horizon of 4- months. In terms of back performance, the CILI has been able to track all the turning points in advance.

6.3. The CILI for the IIP growth rate cycle, has been constructed based on four economic indicators, viz., Non-food credit, WPI- Food Articles, Cement Production and US GDP. The CILI is found to have a forecast horizon of 4- months. In terms of back performance, the CILI has been able to track all the turning points in advance, except the trough observed during June 2001. This turning point was missed by the CILI by one month.

6.4. Alternatively, a CILI was constructed based on the first four principal components of five indicators, viz., Non-food credit, WPI- Food Articles, Cement Production, Inter-Bank Forward Premia of US Dollar (6-months) and US GDP, for the IIP growth rate cycle. The forecast horizon for the CILI is found to be of 4-months. In terms of back performance, the CILI has been able to track all the turning points in advance.

6.5. The Composite Index of Leading Indicators (CILI) for the Non-agriculture GDP (NAGDP) growth cycle, has been constructed based on five leading economic indicators, viz., Currency with the Public, WPI- Manufactured Product, Production of Commercial Motor Vehicles, IIP- electricity and Non-oil Imports.

6.6. Alternatively, a CILI was constructed based on the first three principal components of five indicators, viz., Currency with the Public, WPI- Manufactured Product, Production of Commercial Motor Vehicles, IIP- electricity and Non-oil Imports, for the NAGDP growth cycle. The CILI has a forecast horizon of 2- quarters. In terms of back performance, the CILI has been able to track all the turning points in advance, except the peak observed during 1998-99:Q1. This turning point was captured by the CILI synchronically.

6.7. The CILI for the NAGDP growth rate cycle, has been constructed based on four economic indicators, viz., Non-food Credit, WPI-Food Articles, IIP- Capital Goods and US GDP. The CILI has a forecast horizon of 2- quarters. In terms of back performance, the CILI has been able to track all the turning points in advance, except the trough observed during 2001-02:Q2. This turning point was captured by the CILI synchronically.

6.8. Alternatively, a CILI was constructed based on the first three principal components of five indicators, viz., Non-food Credit, WPI- Food Articles, IIP- Capital Goods, Inter-Bank Forward Premia of US Dollar (6-months) and US GDP for the NAGDP growth rate cycle. The CILI has a forecast horizon of 2- quarters. In terms of back performance, the CILI has been able to track all the turning points in advance, except the trough observed during 2001-02:Q2. This turning point was captured by the CILI synchronically.

6.9. The Composite Index of Coincidental Indicators (CICI) for the monthly IIP growth rate cycle has been constructed based on three economic indicators, viz., IIP- Intermediate Goods, IIP- Consumer Durables and Bombay Stock Price Index- 30 script. In terms of back performance, the CICI has been able to track all the turning points synchronically with the reference series, except the peak observed during March 1996.

6.10. The CICI for the quarterly NAGDP growth rate cycle has been constructed based on three economic indicators, viz., IIP- General Index, Total exports and Total Bank Credit. However, in terms of back performance, the CICI has missed the last three turning points synchronically as that of the NAGDP series.

6.11. For the empirical exercise, the sample covered the period from April 1990 to March 2006. In the month of March 2006, the CILI based on these data, indicated upward movement in the industrial production to continue at least till July 2006. Presently the quick estimate of IIP is available till July 2006, which shows upward movement in the industrial production. This confirms the performance of the CILI.

Section VII
Recommendations of the Group

7.1. The Group examined the recommendations of the earlier Group (Report of the Working Group on Economic Indicators, RBI 2002) and found them very useful. Steps have been taken by the Bank to address some of these recommendations, viz., improving the industrial outlook survey and estimation of capacity utilisation. However, further work is required to strengthen the database on inventories.

7.2. The Group examined the existing literature, both domestic and international, on construction of Leading Economic Indicators. The empirical exercise was carried out using the techniques adopted by the international organisations and also based on the methodologies suggested in some of the research papers. Empirically, both the Growth Cycle and Growth Rate Cycle methodologies have been attempted. For the selection of the leading indicators, both the linear method (through correlation coefficient) and non-linear methods (Self Organising Maps) have been attempted. As Institutional effort, still substantial further work needs to be done to arrive at the final conclusion about the methodology to be adopted. More work needs to be undertaken to examine the applications of Neural Network for unravelling the non-linear relationship among the indicators.

VII.1 Selection of the Reference Series

7.3. The monetary policy review is undertaken at quarterly intervals. In order to provide the inputs on the movements of the economic growth for the quarterly reviews, the reference series should be of the same or higher frequency. Out of the possible economic growth series, the quarterly GDP and monthly IIP series qualify for selection of the reference series.

7.4. The business cycle analysis even for developed countries often excludes agricultural output, in view of the dependence of the agricultural sector on natural factors like weather and the low role of interplay of market forces in determining their level. The agricultural growth in India is still largely subject to vagaries of Monsoon. In view of this, the Group recommended the non-agricultural GDP as the main reference series.

7.5. The selection of these two series was guided by the fact that: a) the non-agricultural GDP is a comprehensive measure of economic growth, and b) more data points were available for the index of industrial production. The information regarding the turning points based on the leading indicators for these two reference series would be more robust as compared to dependence on a single series.

7.7. Understanding the phases of expansions and contractions of economic activity, in order to trigger timely appropriate counter cyclical measures, is of paramount importance to the Reserve Bank of India. In view of this, the information on the turning points need to be made available well in advance as the monetary policy measures impact after a certain lag.

7.8. The Group, therefore, recommended that the lead period should be at least 2 Quarters ahead in the case of quarterly data (i.e. non-agricultural GDP) and 4 months ahead in the case of monthly data (i.e. Index of Industrial Production).

7.8.The Group carried out the empirical exercise for construction of leading indicator applying the appropriate methodology used by the leading international organisations.

7.8.1 The empirical exercise was carried out using growth cycle and growth rate cycle methodologies. The results could not provide sufficient evidence in favour of any one methodology. The Group, therefore, feels that the leading indicators should be constructed using both these methods until sufficient confidence on any particular method is established.

7.8.2. For estimating the relationship between the reference series and the leading indicators, both, linear regression model and non-linear model (probit model) were applied. In this case also the Group held the view that the empirical work should be based on both types of models.

7.8.3. The Group deliberated upon the issue of minimum number of indicators to be considered for constructing the composite leading index. It was observed that most of the earlier work, in this area, had considered at least four indicators for constructing the composite indicators. The other most preferred option, observed from the literature surveyed, had been to construct the composite leading indicators using principal components. The Group recommended that these approaches should be followed for the construction of leading economic indicators.

7.9. The Group underlined the importance of developing institutional capabilities for undertaking leading indicators work. The Group noted that, during the initial stages, the RBI should take a pivotal role in the area of business cycle analysis. Its involvement is necessary to ensure that robust indicators are developed and policy response is based on reliable information base in the field.

7.10. In order to take up this work at RBI, a separate unit should be set up without further delay. The results should be discussed in a forum where the correctness of the results and the methodological issues could be deliberated. In this way the methodology for compilation of index of indicator could become technically strong. RBI may avail of the help of technical experts on a continuous basis till the methodology stabilises. In the course of a year or so, RBI should disseminate review of the business cycle and leading indicators to the public on quarterly basis.

7.11. Since the methodological issues need further strengthening, the Group recommended that a Standing Committee, including experts from the Universities or other institutions and RBI should be set up.

7.12. From the methodological points of view, explorations may also be attempted to estimate the reference series through the application of the dynamic factor model, in the future.

7.13. The Group also felt that for the members of the Reserve Bank associated with the work on Leading Indicators, some international exposure in terms of training, etc., at OECD, ECRI or other such institutions should be provided.

7.14. The Report could be put up in RBI web site for feedback and comments.

7.15. Over the period, RBI should prepare leading indicators for inflation, exchange rate and trade also.

In view of the need for development of a basket of coincident, leading and lagging economic indicators to monitor the movements of the Indian economy on a regular basis based on appropriate methodology and the international best practices, it has been decided to set up a Technical Advisory Group on 'Development of Leading Economic Indicators for Indian Economy'.

2. The Terms of Reference of the Technical Advisory Group are as given below:

ii. To recommend modalities of outsourcing the work for construction of indicators by appropriate external agency or institution, including scope of work and deliverables.

iii. To evaluate the work of the external agency and recommend its acceptance by the Bank.

iv. Any other issue as deemed necessary for development of indicators on business cycle.

3. The constitution of the Technical Advisory Group would be as follows:

Annex II.A.

The HP filter does not rely on any priori information about business cycle peaks and troughs; it is applied mechanically. The filter amplifies the business cycle frequencies and dampens short- and long- run variations. Consider an observed time series Y_t, which is assumed to be the sum of the cyclical component C_t and trend component T_t, i.e., Y_t = C_t + T_t.

The parameter λ reflects the relative variance of the trend component to the cyclical component. Thus λ controls the smoothness of the trend. For a given λ, the HP filter tries to choose the trend component T_t to minimize C_t, so that the following loss function is being minimized,

For λ = 0, there is no penalty on trend adjustment, and the actual observed series becomes the trend component. For large values of λ, the trend component approaches a linear trend. Lower value of λ produces a trend that follows close to the original series. For annual series, generally one takes λ = 100 and for quarterly series λ = 1600. For monthly series, generally one chooses λ = 14400. However, Serletis and Krause (1996) suggested λ to be equal to 129600 for monthly series.

The BP- filter is used to isolate the cyclical component of a time series by specifying a range for its duration. It is a linear filter that takes a two-sided weighted moving average of the data where cycles in a 'band', defined by a specified lower and upper bound, are 'passed' through, and the remaining cycles are 'filtered' out.

Generally, it is expected that the business cycle lasts somewhere from 1.5 to 8 years so that the cycles in this range can be extracted. For quarterly data, generally the range corresponds to a low duration of 6, and an upper duration of 32 quarters.

The probit approach is an alternative method to assess the usefulness of the set of leading indicators in forecasting the turning points (Estrella and Mishkin, 1996; Lamy, 1997; Mensah and Tkacz, 1998; Gaudreault and Lamy, 2002; Krystalogianni et al. 2004). The dependent variable (T) is defined as-

T_t = 1 for a period of recession in period t

= 0 otherwise

Given the set of n leading indicators X = (x1 , x2 ,- - -, xn ), a standard regression model in this case would be,

Where X it ki is the i(th) - ileading indicator at time t with a leading period of k_i months. Equation (---) can be re-written as,

where is the set of coefficients corresponding to the set of indicators. The main objective of using the probit approach is to estimate a response probability,

where F is the normal distribution, given as,

The time domain analysis assigns all frequencies equal weights, whereas the frequency domain analysis is conducted on a frequency-by-frequency basis, using the entire frequency range 0 to . For a univariate series, the method identifies how much of the total variance is determined by each frequency components.

The spectral density is obtained by smoothing the periodogram using lag windows (weights assignments) such as Bartlett's, Tukey's and Parzen's.Bartlett's window assigns linearly decreasing weights to the auto-covariances in the neighborhood of the frequency considered and zero-weights thereafter.

The population Spectrum can be estimated non-parametrically using,

where m is defined as the bandwidth parameter and is generally taken as

The US Conference Board compiles leading, coincidental and lagging indicators for nine countries all over the world, viz.,, the United States, Australia, France, Germany, Japan, Korea, Mexico, Spain and the United Kingdom. The methodology followed is the month-to-month growth rate cycles. The reference series is the GDP.

II.A.1. List of Indicators for the US economy

II.A.2. List of Indicators for Germany

II.A.3. List of Indicators for Japan

Currently the OECD compiles CILI for 29 of its 30 member countries and for six non-members countries. The OECD CLI system is based on the 'growth cycle' approach where cycles and turning points are measured and identified in the deviation-from-trend or ratio-to-trend series. The main reference series used in the OECD CILI system for the majority of countries is industrial production (IIP) covering all industry sectors excluding construction. This series is used because of its cyclical sensitivity and monthly availability while the broad based Gross Domestic Product (GDP) is used to supplement the IIP series for identification of the final reference turning points in the growth cycle.

II.B.1. List of Indicators for the US

1. Dwellings Started
2. Net New Orders For Durable Goods
3. USA CLI Share Prices: NYSE Composite
4. Consumer Sentiment Indicator
5. Weekly Hours of work: Manufacturing
6. Purchasing Managers Index
7. Spread of Interest Rates

II.B.2. List of Indicators for Germany

1. IFO Business Climate Indicator
2. Orders Inflow/Demand: Tendency (Manufacturing)
3. Export Order Books: Level (Manufacturing)
4. Total New Orders (Manufacturing)
5. Finished Goods Stocks: Level (Manufacturing)
6. Spread of Interest Rates (% per annum)

II.B.3. List of Indicators for Japan

1. Inventories to Shipments ratio (Mining And Manufacturing)
2. Ratio Imports to Exports
3. Ratio loans to deposits
4. Monthly overtime hours, Manufacturing
5. Construction: Dwellings started
6. Share Price Index - Tokyo
7. Spread of interest rates

II.B.4. List of Indicators for China

1. Production of Chemical Fertilizer
2. Production of Non Ferrous Metals
3. Cargo Handled at Ports
4. Enterprises Deposits
5. Monetary Aggregate M2
6. Imports from Asia

II.B.5. List of Indicators for Brazil

1. Orders inflow (Manufacturing): Future Tendency
2. Industrial Production of Semi Durable and Non Durable Goods
3. Share Price Index: All Shares
4. Terms of Trade
5. Total Exports F.O.B.

Dua and Banerji (1999), constructed a monthly coincidental index comprising of the following five indicators,

1. GDP at factor cost at constant prices
2. IIP- General Index
3. Wages to workers in factory sector
4. Monthly registered unemployment
5. IIP- consumer goods.

The study proposed the above coincidental index as the reference series to represent the business cycles and growth rate cycles for the Indian economy. The period covers monthly data from 1957 onwards.

Chitre (2001) constructed a monthly reference series for the Indian economy comprising of eleven indicators measured as a deviation from their respective long-term trend. The study covers the period from 1951 to 1982. The list of the indicators is as follows:

The set of leading, coincidental and lagging indicators for peaks and troughs are classified as follows:

Mall (1999) developed a composite leading indicator for the quarterly industrial production based on the growth cycle methodology. The study covers the period 1981:Q1 to 2000:Q1. The IIP-Manufacturing has been taken as the reference series, as the cyclical movement in the IIP-Manufacturing and Non-Agriculture GDP was found to be same. The list of leading indicators used in the study comprise of:

Annex IV.D.

Recently OECD (2006) has developed a composite leading index for the Indian economy. The monthly index of industrial production, excluding construction has been taken as the reference series. The period covered starts from 1995 onwards. The following list presents the leading indicators selected for the Indian economy.

In the context of the Indian economy, the agriculture and allied sector contributes almost 20 per cent to the total GDP. However, till now the agriculture sector depends heavily on the rainfall. The agricultural output declined significantly during the years with deficit rainfall, causing a fall in the overall GDP also.

The following chart represents the quarterly growth rates of the GDP, along with the two sectors, agricultural GDP and non-agricultural GDP, over the period 1991-92:Q1 to 2005-06:Q4. It can be observed that the growth rate of the non-agriculture GDP remains stable during the period. On the other hand, presence of volatility can be easily observed in the agriculture sector growth rates.

During the same period, the standard deviation for the agriculture sector is found to be 5.2, which is significantly higher than 2.4 observed in the non-agriculture sector. The standard deviation in the overall GDP is found to be at 2.3. This prefers the selection of the non-agriculture GDP sector as the reference series compared to the overall GDP.

Chart 11:Growth Rates of the GDP

Note: BEA: Bureau of Economic Analysis, USA; BSE: Bombay Stock Exchange; CMIE: Centre for Monitoring Indian Economy; CSO: Central Statistical Organisation; DGCI&S: Directorate General of Commercial Intelligence & Statistics; IMF: International Monetary Fund; OEA: Office of the Economic Adviser, Ministry of Commerce and Industry, Government of India; OECD: Organisation for Economic Co-operation and Development; RBI: Reserve Bank of India; SEBI: Securities and Exchange Board of India.

Annex VII.A.

Dynamic Cross Corelation of Monthly IIP with Indicators - Growth Cycles - HP filter (129600)

Annex VII.B.

Dynamic Cross Correlation of the Indicators with Non-Agriculture GDP (Quarterly data)
Growth Cycles

Annex VII.C.

Dynamic Cross Correlation of Monthly IIP with Indicators - Growth Rate Cycles

Annex VII.D.

Dynamic Cross Correlation of the Indicators with Non-Agriculture GDP (Quarterly data) - Growth Rate Cycles

Annex VIII.A.

Clustering of the Quarterly Indicators based on SOM

Note: Figures presents the deviations from the overall mean.

Note: Figures presents the deviations from the overall mean.

1. Indian business cycles through the nineties are non-uniform. The variables isolated in the indicators differ from earlier studies. There is more representation of financial variables. Since change is rapid, particularly in the last few years, it may be worthwhile to redo the identification exercise of leading indicators for the IIP growth rate cycle, focusing on the period 2000 onwards.

2. The de-trending exercise should also be done separately for the latter period. Even polynomial trends are sensitive to the period used. A graph should show what the business cycle looks like if the trend is estimated over the later period.

Response to the points 1 and 2:

The Indian economy experienced only two turning points after 2000, based on the two methodologies followed in the study. Identification of leading economic indicators based on only two turning points may not be ideal. In case the analysis is carried out based on the period 2000 onwards, we will have only 72 monthly and 24 quarterly observations, which would be too few to carry out a meaningful business cycle analysis. Accordingly, this exercise will be carried out over time as the number of observations becomes sufficiently large.

3. An effort should be made to prepare time series on term structure of interest rate. Similarly a time series on housing credit can be prepared to substitute for housing starts.

Response:

As suggested efforts would be made to prepare time series data on the above mentioned indicators.

Artis, M.; Bladen- Hovell, R.C.; Osborn, D.R.; Smith, G. and Zhang, W. (1995), 'Predicting turning points in the U.K. Inflation Cycle', The Economic Journal, Vol. 105, 1145- 1164.

Bikker, J.A. and Kennedy, N.O. (1999), 'Composite Leading Indicators of Underlying Inflation of Seven EU Countries', Journal of Forecasting, Vol.18, 225- 258.

Birchenhall, C.R., Jessan, Osborn, D.R. and Simpson, C. (1999), 'Predicting U.S. Business Cycle Regimes', Journal of Business and Economic Statistics, Vol. 17(3), 313- 323.

Bry, G. and Boschan, C. (1971), 'Cyclical Analysis of Time Series: Selected Procedures and Computer Programs', National Bureau of Economic Research, Technical Paper No. 20.

Burns, A.F. and Mitchell, W.C. (1946), Measuring Business Cycles, National Bureau of Economic Research, New York.