Hamilton, J., 1989, “A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle,” Econometrica 57.
Mongardini, J. and T. Saadi-Sedik, 2003, “Estimating Indexes of Coincident and Leading Indicators: An Application to Jordan,” IMF Working Paper WP/03/170.
Simone, A., 2001, “In Search of Coincident and Leading Indicators of Economic Activity in Argentina,” IMF Working Paper WP/01/30.
Stock, J. and M. Watson, 1989, “New Indexes of Coincident and Leading Economic Indicators,” NBER Macroeconomic Annual Report, pp. 351-394.
Stock, J. and M. Watson, 1991, “A Probability Model of the Coincident Economic Indicators,” Leading economic indicators: new approaches and forecasting records, edited by K. Lahiri and G. Moore, Cambridge University Press.
Stock, J. and M. Watson, 2002, “Forecasting Using Principal Components from a Large Number of Predictors,” Journal of the American Statistical Association, December 2002, Vol. 97, No.460.
Prepared by Hiroko Oura (EP, APD).
While revenue and expenditure data are available on a monthly basis for the central government, statistics measuring local government activities have a lag of two years.
Real sector and trade data are chosen from CEIC. Data are picked if they are available (1) on monthly basis and (2) with a few months lag (usually 2–3 months).
They included the U.S. and Euro area producer price indices and composite leading indicators from the OECD.
As discussed later in the paper, nominal series often produce more stable estimates than real series, which are often affected by noise in their deflator. For instance, the artificial spike in headline CPI in 4Q 2005 owing to domestic fuel price hikes distorted the deflated real series. In addition, volume data for trade were not used because: (1) import and export deflator estimation in Indonesia is known to be weak; and (2) there are longer lags with respect to the release of data on trade volumes.
Principal component estimates using 54 candidate variables (Stock and Watson, 2002, used 130), showed that more than 9 principal components would have to be included with data available only for 20 quarters this would not give meaningful results.
One benefit of the model-based approach over earlier “ad hoc” studies is to let the econometric model decide which indicators should be combined into a composite index and with what weights. However, more recent studies have found that model based composite indices are very similar to the equal weighted ones (Marcellino (2005)), and that the estimation of economic conditions are rather robust to the choice of method. In addition, estimated optimal weights might be strongly influenced by in-sample developments, especially when the sample is small, and produce relatively poor out-of-sample forecasts. Therefore, alternative indices are analyzed, some with optimal weights and others with equal weights, and compare their performances with our data.
Nominal series are preferred over real series, because the estimation results are strongly influenced by the jump in the headline CPI in the fourth quarter of 2005, owing to a large domestic fuel price hike and subsequent weakening in consumption in the quarter.
The estimated coefficients in a multivariable regression provide weights on each included indicator in a straightforward manner. For instance, with model 6, an optimal weight composite index is constructed by adding standardized consumer credit growth and currency in circulation growth after each of them are multiplied by their respective estimated coefficients.
Indeed, serious multicolinearity problem appeared with Indonesian data in preliminary analysis in which efforts were made to estimate optimal weight models following Mongardini and Saadi-Sedik (2003), by starting from a general model including majority of the variables, and eliminating one by one based upon the t-statistics of coefficients. Using principal components for estimation instead of raw indicators is one way to avoid this multicolinerity issue; however, as discussed in section C, a relatively large number of estimated principal components are needed to capture a reasonable portion of overall variations with Indonesian data.
The importance of capital goods imports as an indicator of investment was confirmed during a series of meetings with the authorities and the private sector participants. The strong relationship between capital goods import and investment was detected by varieties of methodologies, including simple correlation analysis and the OECD methodology (BI), with different data clean-up procedures.
It may appear implausible for motor vehicle sales to be a leading indicator for investment when it is not strongly related to private consumption. However, a large portion of local motor vehicle sales are of commercial, not passenger, vehicles. This means that car sales may reflect business demand more than private consumption.
Comparators include South Korea, Malaysia, the Philippines, and Thailand. The numbers are based on 2000–2005 average.