Appendix II. Data Sources and Definitions
- Robert Corker, and Wanda Tseng
- Published Date:
- March 1991
While theoretical considerations suggest those variables most likely to influence money demand, available data do not conform to “ideal” economic concepts. In particular, most monetary aggregates incorporate banking sector liabilities of varying degrees of liquidity, holdings of which may be influenced by a variety of motives. Some of these liabilities may not perform the role of “money” in the economy very well, while other nonbank liquid liabilities, which are excluded from the aggregate, may be more appropriate components.53 The definitions of income, prices, and interest rates are similarly subject to measurement problems.
Prior to analysis, money and income data were seasonally adjusted using the XII filter. While some econometricians (e.g., Wallis (1974)) have argued that such pre-filtering can induce spurious short-run dynamic relationships between the data, seasonal adjustment was carried out for two reasons. First, a lack of data for some series (notably interest rates) limited the degrees of freedom available for the econometric analysis in many cases: the use of seasonal dummies in the regressions would have exacerbated this problem. Second, for many countries, quarterly income data were interpolated from annual observations implying pre-fittering of some of the variables anyway.
An analysis of the time series properties of the data showed that most series were integrated of order one.54 That is, most of the data fluctuate randomly around a persistent trend. No series was found to be integrated of order two—that is, to fluctuate around an accelerating trend—although a few series were integrated of order zero and are therefore stationary. Most interest rate series were integrated of order one, although this latter result may reflect the size of the data sample: over longer periods, interest rates are probably mean reverting. Similarly, the income velocities of both broad and narrow money were generally integrated of order one (narrow money in Malaysia and Indonesia were exceptions) indicating the presence of significant long-run trends in the 1970s and 1980s. The following describes the specific data definitions used in the empirical analysis.
Both narrow money—currency plus demand deposits—and broad money—narrow money plus time and savings deposits (quasi-money)—were analyzed separately. The definitions of money correspond closely, in most cases, to the aggregates M1 and M2. An average of beginning and end-period money stocks was used in the regression analysis.
Data for narrow and quasi-money were taken from lines 34 and 35, respectively, of the International Monetary Fund’s International Financial Statistics (IFS). For Indonesia, narrow and broad aggregates were adjusted down by Rp 400 billion from the second quarter of 1989 onward to adjust for step increases owing to the classification of a new financial institution in the banking survey. Data for Malaysia were taken from the Monthly Statistical Bulletin (Bank Negara Malaysia) and incorporate revised estimates of M1 and M2 that include repurchase agreements.
Income and Prices
Income was proxied by GDP or GNP, as available, and prices were proxied by the GDP (GNP) deflator. Data for real and nominal GDP (GNP) were taken from IFS, lines 99b.p (99a.p) and 99b (99a), respectively, except for data for Korea and Singapore, which were taken from the national accounts yearbooks of these countries. For Malaysia and Indonesia, quarterly data for real GDP were interpolated from annual observations according to the pattern of quarterly movements in industrial production (Malaysia) and a weighted average of industrial and oil output (Indonesia). Quarterly data for real GNP for the Philippines were interpolated from semiannual data for the period 1973–79 using the pattern of quarterly movements in industrial production. Quarterly data for real GDP in Myanmar, Nepal, Sri Lanka, and Thailand were interpolated from annual data using a cubic spline function. In the case of Nepal, the data were first converted to a calendar year basis from the reported fiscal year basis. Quarterly data for the GDP (GNP) deflators in Indonesia, Myanmar, Nepal, the Philippines, Sri Lanka, and Thailand were interpolated from annual data according to quarterly movements in the consumer price index: for Malaysia, the patterning series was a weighted average of the consumer price index and export unit values.
For broad money, market interest rates on alternative assets were generally proxied by either money market, treasury bill, or international interest rates (such as the London interbank offered rate—LIBOR).55 The rate of return on broad money was approximated by a representative deposit rate multiplied by the share of quasi-money in broad money. This definition assumes that all components of narrow money are non-interest bearing. For narrow money, the return on alternative assets was proxied by deposit, money market, or price inflation rates. In principle, a vector of alternative returns should enter the money demand functions. In practice, multicollinearity between interest rates implies that only one rate can feasibly be used to measure opportunity cost. Some degree of judgment was used to determine the “best” proxy in specific cases.
Data for money market, treasury bill, and deposit rates were taken from (where available)IFS lines 60b. 60c, and 601, respectively. Data for the three-month (LIBOR) were taken from line 111601dd of IFS. In Tables 1 and 2, the interest rate variables for narrow money are all deposit rates. For broad money, the interest rate variables are the following: Indonesia and Malaysia, three-month LIBOR minus deposit rates weighted by the share of quasi-money in broad money: Korea, the interest rate in curb markets (provided by the Korean authorities) minus deposit rates weighted by the share of quasi-money in broad money; the Philippines. 91-day treasury bill rate minus deposit rates weighted by the share of quasi-money in broad money; Singapore, the money market rate; Sri Lanka, deposit rates weighted by the share of quasi-money in broad money; and Thailand, money market rates minus deposit rates weighted by the share of quasi-money in broad money.