Composite Coincident and Leading Economic Indexes

Composite Coincident and Leading Economic Indexes This article presents the method of construction of the Coincident Economic Index (CEI) and Leading Economic Index (LEI) and the use of the indices as a tool to predict future economic activity and its turning points. The article also provides guidelines for the use of LEI to forecast quarterly and monthly * GDP, known as indicator approach to forecast GDP growth. This approach has been applied in many countries, even though it is considered as a forecast without theory 1. The article comprises 5 sections. Section 1 gives detail of selection of indicators for constructing CEI and LEI indexes. In Section 2, the method of construction of CEI and LEI and their results are described. In Section 3, we compare CEI and LEI constructed by the Bank of Thailand and those constructed by the National Economic and Social Development Board and the Ministry of Commerce. Section 4 describes application of CEI and LEI in the forecasting of quarterly GDP. Brief concluding remarks are in Section 5. Section 1: Selection of indicators for constructing CEI and LEI CEI is a wavelike index having the turning points with its up and down swing (expansion and contraction) move in accordance with the overall economic activity. LEI is a wavelike index which its turning points lead the turning points of economic activity. In other words, the turning points of LEI lead the turning points of CEI. The countries, having monthly GDP or large coverage of the Manufacturing Production Index (MPI), can use them as references to indicate which variable is coincident or leading indicator when constructing the composite coincident and leading economic indexes. For Thailand, monthly GDP is not available. In addition, value of production of industries included in the MPI represents merely 32 percent of GDP * This is the author s academic point of view and not The Bank of Thailand s point of view. 1 This is according to Auerbach J. Alan, as quoted from The Index of Leading Indicator: Measurement without Theory, The Review of Economics and Statistics 1982. Moreover, LEI can be used to forecast other economic variables for instance the University of Illinois uses it to forecast the US. Exchange rate in comparison with that of Canada.

2 and the MPI series is not long, beginning from 1985. Therefore, selection of variables (or components) to construct CEI and LEI is justified from economic theory and research findings, more than from identified turning point dates of component series compared with those of the MPI. 1.1 Variables used to construct the composite CEI include:! Department stores sales! Imports volume! Debits to demand deposit! Car sales (passenger and commercial cars)! Manufacturing Production Index (MPI)! Business tax, and Value Added Tax or VAT (adjusted to 7 percent rate). The selection of variables for constructing CEI is mainly based on economic significance. Hence, these variables should be correlated with economic activities. Yet, there is a controversial argument that imports should be more suitable variable in the construction of LEI than CEI. But there are several studies claiming that imports are coincident variable. This is because increased imports are the result of increased income or increased economic activities. In our work, imports are used as coincident indicator. In case of VAT, submission of VAT is usually one month delayed. Therefore, VAT that reported by the Ministry of Finance in the current month reflects economic activities of the past month. Therefore, in constructing CEI, say, of the month of April, VAT reported in May is chosen as our database. 1.2 Variables used to construct LEI are:! Permitted construction areas! Capital of newly registered companies! Broad money including finance companies and finance and securities companies (M2A)! Number of foreign tourists! Stock price index! Exports volume! Crude oil price index in the Oman market (invert) Selection of variables to construct LEI is also based on the economic significance. In constructing LEI, turning points (TP) of the seven mentioned variables are tested against CEI's. Exports are regarded as leading indicator. This is because increased exports lead to good

3 expectation of production, and thereby resulting in increased economic activities in latter periods. Apart from the economics significance and the TP, selection of variables are also dependent on: * Market expectation such as stock price index, crude oil prices and exports * Prime mover such as M2A which reflects liquidity in the economic system * Production time. This is due to the reason that it takes time for any real economic activities to take place. Good examples are permitted construction areas and capital of newly registered companies. * External influence such as exports and the number of foreign tourists. * Data availability. Monthly data must be the series that are long enough to construct indexes and conduct TP test. 2. Calculating composite CEI and LEI ( characteristics and results of the two indexes) 2.1 The procedure for calculating CEI and LEI CEI and LEI are constructed by employing growth cycle approach rather than classical approach. (Growth cycles measure fluctuations in growth rates of economic activity, while classical cycles measure ups and downs in the level of economic activity). After World War II, many countries including Thailand have continued economic growth. In other words, they have been affected more by growth recession than by the reduction of economic activity levels. Therefore, growth cycle approach is more appropriate. Growth cycle approach compares growth rate of the economy with the long-term growth rate trend. If the overall expansion of the economic activity during a certain time is lower than the long-term trend, growth recession is taking place irrespective of the levels of the economic activity. On the contrary, if the overall expansion of the economic activity is higher than the long-term trend, growth expansion is taking place. Nevertheless, in some countries if economies generate fluctuations in the level of economic activity, classical cycle approach may be appropriate than growth cycle approach.

4 Components of CEI! Department store sales! Imports volume! Debits to demand deposit! Sales of automobiles! Manufacturing production index! Value added tax (7 % rate) Components of LEI! Permitted construction areas! Capital of newly registered companies! M2A! Number of foreign tourists! Exports volume! Crude oil prices index in Oman Market ( invert ) The calculating procedure of CEI and LEI 1. Remove seasonal factor and price effects from every component series (X i,t ), where as t = month, i = Indicator. X i,t = Seasonally adjusted series component i in month t in real term. 2. Compute the symmetric percentage change of the real seasonally adjusted data x i, t = X i, t X i, t X i, t + X i, t 2 1 1 * 100 t = month, i = Indicator 3. Standardise X i,t, using the following formula Where S i, t = x i, t A. A N t 2 = = N x i, t 1

4. Calculate the (symmetric) percentage change in index (R t ) 5 R t = 10 W i = 1 10 i = 1 i S i, W i t 5. Standardise R t to adjust its amplitude consistent to the trend-cycle component of coincident index r t = R t / F where F = N t = 2 N t = 2 R P t t P t = Coincident Index of R t, hence in the case of calculating composite coincident index, F equals 1. 6. Calculate Cumulative Index by; I t = I t-1 * (200+ r t ) (200-r t ) Note: The first month I t-1 is set to be equals 100. Then, the rebase of I t [e.g. I t (1990) = 100] is implemented. 2.2 The indicating procedure of Cycles and Turning Points by using Bry- Boschan program Theoretically, in order to indicate cycle and turning points, Bry- Boschan program try to delete Trend and Irregular components from the seasonal adjusted data (i.e. will only left cycle in the data). The program procedure is as follow. 1. Calculate the Centred 75-month moving average. This is to delete the Irregular variation to get the first trend. The implication of 75-month as the calculation period is due to the reason that the cycle duration is

6 not exceed 6 years. In some countries, e.g. South Korea and Thailand, the cycle duration is expected to be about 60-65 months. Thus, the indication of number of months in the moving average procedure is adjustable. 2. Roughly calculate the standard deviation of the first trend and find TPs and Phases (From the first turning point to the next) 3. From 2., calculate phase averages of each data in each phase 4. Calculate the 3-Phase moving average of the figures from 3. 5. Calculate the trend line, by joining the middle point of the 3-Phase moving average. This is to get the slope or the Phase average trend 6. Smooth the Phase average trend by using the 12-month moving average. The calculating of the trend line using the Phase average trend procedure is better than other calculation methods in the sense that the Phase average trend procedure can eliminate the influence of short-term cyclical movement. Note that the influence of short-term cyclical movement occurred by the different in time period of each cycle. During the indication of cycle and TPs, Program is set accordingly by cycle time and Phase, i.e. each cyclical movement should not be less than 15 months, and each Phase should not be less than 6 months. After the acquiring of the data s phase average trend and cycles line, the program will indicate Peaks, Troughs, Expansion and Contraction Phase. The distortion of the time-series data from the Phase average trend will show the data cyclical movement, i.e. the part below the trend will show the Growth Contraction, on the other hand, the part above the trend will show the Growth Expansion. 2.3 Characteristics of CEI and LEI constructed by the Bank of Thailand In constructing CEI and LEI based on the monthly data of the above components by using the computation methods of the programs developed by The Foundation for International Business and Economics Research (FIBER), University of Columbia as presented in 2.1, and by using the Bry-Boscham program in determining the cycles and TPs as presented in 2.2, it is found that from 1978 (2521 Buddist Era) to 1999 (2542 B.E) (April) Thailand has passed three and a half economic cycles, with each cycle lasting about 58 months. The duration from the peak to the next trough (recession phase) averaged 25 months, and the duration from the trough to the next peak averaged 33 months. The peaks and troughs of CEI and LEI computed by Bry-Boschan program can be presented as below:

7 LEI CEI Duration (months) Troughs 2524:09 2529:05 2535:05 2541:05 2525:01 2529:06 2535:05 2541:08 4 6 0 3 Average lead time 3.25 at trough Peaks 2521:11 2526:09 2533:06 2539:02 2522:10 2527:01 2533:07 2539:07 11 4 1 5 5.25 Average lead time at peak Average lead time 4.25 From the CEI, the trough of the latest cycle fell in August 1998(B.E.2541), and the duration from the trough to the next peak (expansion phase) during the recovery period was 33 months. Thailand is therefore believed to have passed the trough and is within the recovery period. The results of CEI and LEI are presented in the illustration 1, and changes of indicators in table 1. 3. Comparison of CEI and LEI conducted by the Bank of Thailand, Office of the National Economic and Social Development Board and Ministry of Commerce. Indices Economis Indicators Team, Bank of Thailand CEI Components 1. Department stores sales 2. Imports volume 3. Debits to demand deposit 4. Car sales ( including trucks) 5. Manufacturing production index VAT ( 7 percent)

8 LEI 1. Permitted construction areas 2. Capital of newly registered companies 3. M2A 4. Number of foreign tourists 5. Stock price index 6. Exports volume 7. Crude oil prices index in the Oman market (invert) Ministry of Commerce CEI LEI 1. Car production volume 2. Motorcycle production volume 3. Beer production volume 4. Car sales 5. Imports volume 6. Customs 7. Department stores sales 8. VAT 9. Cement production volume 1. Registered capital for new businesses 2. Permitted construction areas 3. Number of foreign tourists 4. Stock price index 5. Narrow money ( M1) 6. Exports volume National Economic and Social Development Board CEI LEI The regression analysis method to conduct monthly GDP is implemented. During 1980-1986, dependent variables were government expenditure, indirect tax, and cement output. However, after 1987, the variables used have been the government expenditure, electricity consumption in the private sector, and Manufacturing Production Index. 1. Registered capital of new businesses 2. Permitted construction area 3. Number of foreign tourists 4. Stock price index 5. Narrow money (M1) 6. Crude oil prices index in the Oman market 7. Capital investment of the operating and expanding firms receiving BOI support

3.2 Comparison of economic indexes constructed by the Bank of Thailand, Ministry of Commerce and Office of the National Economic and Social Development Board. 9 Economic forecasts obtained from the indices conducted by the Bank of Thailand, Ministry of Commerce and Office of the National Economic and Social Development Board have all been pointing to the same direction. This means that Thailand has passed the trough and is within the recovery period (how soon Thailand can emerge from this stage depends on how effective the economic measures the government has implemented). However, the lead time of LEIs of these three organizations are slightly different. This is due partly to differences in the selected components and the construction method of CEI. In constructing LEI, the Bank of Thailand, Ministry of Commerce and Office of the National Economic and Social Development Board all use the same technique, i.e. growth cycle technique developed by the Foundation for International Business and Economic Research (FIBER). However, the components used are slightly different. Construction of CEI by the Bank of Thailand and Ministry of Commerce is based on the same principles, those of FIBER. Unlike these two organizations, to obtain CEI, the Office of the National Economic and Social Development Board relies on the regression analysis method. This means that the assessment of monthly GDP from monthly data of variables, which are correlated with GDP, is implemented (see 3.1). Different components and constructing methods of CEI result in difference in the length of lead-time. For example, in the case of the Bank of Thailand, latest data fell in April 1999, but in the cases of the Ministry of Commerce and Office of the National Economic and Social Development Board, the latest data fell in March 1999. The Bank of Thailand Ministry of Commerce Office of the National Economic and Social Development Board Average lead time 5.25 months 5.6 months 7 months at peak Average lead time 3.25 months 1.75 months 4 months at trough Average lead time 4.25 months 3.7 months 5.5 months

10 Peaks and troughs of CEI of the three organizations are presented below: Bank of Thailand P P P P T T T T 79/11 82/1 84/1 86/6 90/8 92/5 96/8 98/8 Ministry of Commerce P P P P P P T T T T T 70/7 72/5 73/11 74/11 79/5 82/2 84/1 86/6 90/8 92/1 95/1 The Office of the National Economic and Social Development Board P P P P T T T 81/4 81/11 83/5 85/8 90/11 93/3 97/3 Differences of the number of cycle (P-T-P or T-P-T) and duration of peaks and troughs lie in the number of months from which data is derived. For instance, the Bank of Thailand had the data up to the month of April 1999, while Ministry of Commerce and Office of the National Economic and Social Development Board (NESDB) held the data up to March 1999. Another factor contributing to the differences is the altered constructing methods of CEI used by NESDB.

11 Average time length of phase and cycle is shown below: The Bank of Thailand Ministry of Commerce Office of the National Economic and Social Development Board Contraction phase 25 months 23 months 18 months Expansion phase 33 months 36 months 43 months Cycle 58 months 59 months 61 months 3.3 Results of LEI of the three organizations In March 1999, LEIs of Ministry of Commerce, Office of the National Economic and Social Development Board and the Bank of Thailand headed towards the same direction, recording a growth from February by 2.3, 0.6 and 0.2 percent, respectively. This leads to the belief that in the next 3-4 months, Thai economy is still within the recovery growth path. The fact that the Bank of Thailand received data quicker than the other two organizations enabled it to compute LEI of April which turned out to stand at 118.1, an expansion from the previous month by 1.0 percent, leading to the projected economic growth in July-August. 4. The use of LEI as a forecasting tool 4.1 Assessment of GDP through index approach Since 1930 s, the United States and the European countries have been widely using leading, coincident and lagging indices as a tool to forecast economy and identify economic policy perspective. In this paper, we have presented the construction and the use of CEI and LEI as a tool to signal economic conditions in section 1-3. For section 4, we have experimented on the relations between LEI and GDP by implementing the assessment of quarterly GDP (QGDP). Assessment of GDP through LEI is considered as an approach through the use of index apart from using econometric model. Unlike the econometric models forecasting approach, the index forecasting approach has not based on economic behavior, or assumption of exogenous variables. However, it depended on the historical data tested and proved to be coincident, leading or lagging prior to constructing of the composite indices. These composite indices reflect

the overall economic behavior and their cycles appear to be consistent with economic cycles. One drawback of the index approach is that it does not reflect the feedback of the variables used as does the econometric models approach. 12 Many researches have experimented on the use of leading index in assessing GDP, manufacturing production, employment and exchange rate volatility. By doing this, leading index will be constructed from the indicators that can reflect movements of the incidents needed to be forecast. In this article, we have used LEI to assess quarterly and monthly GDP basing on the regression technique. The results are summarized below: Regression technique can be diversified into several models, two of which are exemplified below 1. Year-to-year percentage change of real GDP= f (6-month smoothed growth rates of LEI and CEI in the month of the previous year which yielded the best fitted value). 6-month smoothed growth rate is the ratio of current month index to 12 preceding months converted into compound annual rate 2 / In the case of Thailand, 6-month smoothed growth rates of May are used as explanatory variable. This due to the reason that the May growth rates yield better adjusted R 2 and t-statistics than other months' rate. However when assessing quarterly GDP by using regression equation, it is found that the results are rather inaccurate. This implies that the equation may be better used with long leading index (the lead time is about 1 year), the author then decided to use model 2 3 as described below: 2. Regression model: log (quarterly real GDP of 1988) = f{log[ quarterly LEI (-1) of 1988]} 2 6 month smooth growth rate is the growth of x from 12-month moving average (12-MA) of x during 6.5/12 years as computed by [{ (x/12-ma of x) 12/6.5 }-1]*100 3 In doing the assessment of monthly GDP, the author has tried out many ways of GDP regression analysis with LEI and accepted that the assessment yielding the most accurate results of annual and monthly fitted GDP is the one conducted by Office of the National Economic and Social Development Board

13 In this procedure, the coefficient of quarterly LEI(-1) (QLEI(-1)) is estimated using regression technique. We used one quarter (3 months) lag of LEI because, by turning point determination specified in subsection 3.2, LEI leads CEI (or economic activity) by about 4 months. For example, LEI April can forecast the GDP August. LEI is the index that constructed using the Seasonally Adjusted data. Thus, in order to find the relationship between quarterly LEI (QLEI) and quarterly real GDP (QGDP), we use the Seasonally Adjusted QGDP data computed by NESDB as our data. Because this data is real value at 1988 prices, we therefore rebased our LEI from 1990 to 1988. The regression result which show the relationship of QGDP and QLEI is as followed: LOG(QGDPSA88) = 7.152 + 1.267 LOG(QLEI88(-1)) (eq. 1) (5.45) (4.83) Data sample: 1993:01-1999:01 Adj. R 2 = 0.938 DW = 2.06 where QGDPSA88 = Quarterly real GDP at 1988 prices QLEI88 = Quarterly Leading Index revised to 1988 base year, calculated by averaged sum of monthly Leading Index. t-statistics are in parenthesis Seasonally adjusted GDP computed from equation 1 From the above equation, elasticity of GDP with respect to LEI is equal to 1.2767. To project GDP of the current month, we multiply this elasticity to the percentage change of LEI of the last 4 months (LEI(-4)). This follows our finding that LEI leads CEI or economic activity (GDP) by about 4 months. Because there is no actual monthly GDP figure, we decompose quarterly GDP reported by NESDB to obtain monthly GDP. To decompose quarterly GDP, we simply do it by dividing quarterly GDP figure by 3. That is, QGDP of 1999:01 equals 708502, hence average monthly GDP in this quarter is 236167 (708502/3 = 236167). We use

14 this computed monthly GDP as our base value to project GDP in the following months. For example, to forecast GDP in April 1999, we compute the growth rate of GDP in April from March 1999 by multiplying the month-on-month percentage change in LEI in December 1998 (4 months lead) by the elasticity. Then, we multiply this GDP growth to our GDP base value (GDP in March) and add this additional value of GDP to GDP in March. This process will lead to the seasonally adjusted GDP April at 1988 prices (Table 2). The computation of GDP for the months of May-September follows the same procedures. Then, quarterly GDP figures are derived by adding the above projected monthly GDP figures. Non-seasonally adjusted GDP To obtain non-seasonally adjusted GDP, the estimated figure of seasonally adjusted quarterly GDP is multiplied by seasonal factors. Seasonal factors in each quarter (e.g. quarters 2 and 3 of 1999) are average figures of seasonal factors (in quarters 2 and 3) during the years 1997-1998. Figures of seasonal factors are derived from the following formula: Seasonal factor (quarter n th ) = Non-seasonally adjusted GDP ( n th quarter) obtained from NESDB Seasonally adjusted GDP (n th quarter) obtained from NESDB The figures of seasonal factors, the GDP estimated by LEI in quarter 2 nd and 3 rd of 1999 and the quarterly GDP conducted by the Office of the National Economic and Social Development Board, are illustrated in table 2 and figure 2 (the disjointed line represents figures estimated by LEI). 4.2 Comparison of quarterly GDP estimated by LEI and GDP conducted by the Office of the National Economic and Social Development Board (NESDB) Quarterly GDP (QGDP) computed from QLEI using equation 1 is somewhat deviated from quarterly GDP of NESDB. A glance at Table 3 shows that the constructed QGDP was consistent with QGDP of NESDB approximately by 88 percent (consistency achieved is 21 quarters out of 24 quarters). However, its changing rate is somewhat different from that of QGDP conducted by NESDB. Therefore, great caution must be taken in using LEI to forecast GDP. It should be noted here that assessment of GDP is not the aim of this article, it only attempts to exemplify the

15 application of LEI as a forecasting tool. Thus, simple regression model is presented for the ease of analytical process. The estimation of monthly GDP by using LEI (as presented in 4.1) can be achieved by calculating the changing rate of GDP in each month from the figure of elasticity (1.2767). The result then is multiplied by GDP of March 1999 obtained by NESDB (this GDP is derived from the figure of the first quarter of 1999 divided by 3). In this case, we do not estimate GDP directly from equation 1 to avoid inaccuracy of the computation base when computing GDP of the following months. The reason is because GDP of the first quarter, 1999 obtained by using LEI in equation 1 is equal to 702479.1, which is very much different from that of NESDB (708502). Fitted value of GDP computed from LEI in equation 1 is shown in table 3. 5 Conclusion This article is aimed to point out some aspect of the usefulness of CEI and LEI. Not only are they indicators signaling the economic conditions, they are also used to forecast the economy. This article uses the growth cycle approach in constructing the two indexes and comes up with cycles and turning points. Then the leading indexes are used for the assessment of monthly and quarterly GDP. Nevertheless, accuracy of this instrument is dependent on a number of factors associated with the data obtained and the forecasting techniques used. Construction and the use of indexes in the future is believed to be further developed and upgraded in several aspects such as: 1) The indicators used to construct these indexes should be adjusted and modified as necessary as the same indicators may not always provide correct signals. In addition, to achieve the most accurate results new indicators should be tried out when necessary. 2) Each indicator should be given their own specific weight based on the level of their importance. The current study of index construction is still limited as the existing indicators all have the same weight. 3) Techniques in defining the correlation between LEI and variables should be constantly developed to achieve more accurate assessment. Construction and the application of indices are still restricted in some ways. Although it is the first step of the index construction and the economic forecasting, given adequate and constant development, it will in the long run bring innumerable benefits to the country as a whole.

16 Table 1: Variable data of seasonally adjusted leading index and Leading Economic Index (LEI) Percentage change from previous quarter or month (%) 1998 1999 Oct Nov. Dec. Q.4 Jan Feb March Q.1 Apr. Permitted areas Registered capital for new businesses Stock price index Total exports volume Number of tourist arrivals Real M2A Crude oil price index (Oman) Leading Economic Index (LEI) -26.2 218.6 41.5 0.6-0.6 0.4-9.1 4.7-72.5 7.7-2.3-1.3-0.01-1.7-8.6 34.8-3.9 5.9 0.4-0.6-15.8-15.3 72.2 43.7-1.3 2.1 1.6-15.3-19.2-24.3-5.5-4.6 6.8-0.5 17.2 29.4-36.4-1.5 3.0-6.2 0.6-2.0 1.1 15.9-2.1 2.3-1.7 0.5 21.5-7.3-59.5-7.3 1.0 1.7-0.4 9.3-25.6-20.9 30.4 7.2 1.2 0.6 17.5 2.5-1.1 0.08 2.4-0.8-0.03 0.2-1.2 1.0 Table 1(cont.): Variable data of seasonally adjusted coincident index and Coincident Economic Index (CEI) Percentage change from previous quarter or month (%) 1998 1999 Oct Nov. Dec. Q.4 Jan Feb March Q.1 Apr. Department store sales Trade tax and VAT* Total imports volume Debits to demand deposits Manufacturing Production Index Car sales ( including truck) Coincident Economic Index (CEI) -0.2-2.6-0.3-2.6 3.3 18.6-24.7 7.8 2.4 0.2-3.0 3.7 28.1 1.1 3.8-8.8 2.1 4.4-6.1-1.5 3.2-6.9 1.6 35 160.9 11.6-7.5 3.9 0.8-26.3-19.6-2.8 16.1-1.4 0.8 18.6-5.6-13.4 6.5 6.2 0.2-2.3 137.2 7.9 8.2-1.2 2.4-14.4 22.1 20.7 1.5-1.2 3.1 0.7-0.8 1.5 0.6 2.7 0.2-0.3 3.1 1.9 * VAT ( adjusted to 7percent) and reflects real economic activities of that particular month. 6.1