Relative importance of scheduled macroeconomic news for stock market investors

Introduction

This study investigates the relative importance of macroeconomic news releases for stock valuation. Many market participants believe that macroeconomic news has a major impact on the prices of financial assets. The evolution in recent years of an industry devoted to predicting the figures to be released in upcoming releases supports this belief. Over the years, researchers have examined the significance of the information content of scheduled announcements in pricing financial assets in different countries.1 These researchers have focused on explaining the economics of price variations of financial assets following scheduled announcements. Furthermore, the response of various measures of volatility to these scheduled macroeconomic announcements has also been examined in the literature.2 The primary focus of these researchers has been on the magnitude and speed of price movements of financial assets.

A considerable number of scheduled macroeconomic announcements can be regarded as valuable for investors. Macroeconomic announcements in the U.S.A. are regularly covered in the "Week Ahead" section of Business Week. The economic calendar on Yahoo's pages that documents the days and times of different types of impending announcements ranks announcements in order of importance. The exact mode of classification, however, is not revealed at the source.3 Similarly, the Bureau of Labor Statistics (BLS) classifications of major economic indicators indicate the importance of some macroeconomic announcements. These suggest that, among other things, there may be some general economic indicators whose announcements are viewed as more important relative to others and, thus, closely followed by market participants.

This paper examines the significance of 11 scheduled macroeconomic news releases on stock valuation. That is, we differentiate among these announcements and document empirically, in order of significance, their relative importance to stock valuation. The macroeconomic announcements considered here are the Consumer Price Index (CPI), Employment Report (ER), Employment Cost Index (ECI), Producer Price Index (PPI), Productivity and Cost Measure (PCM), Real Earnings (RE), U.S. Import and Export Price Indexes (USIEX), Retail Sales (RS), Gross Domestic Product (GDP), and National Association of Purchasing Managers Survey (NAPM), both manufacturing and non-manufacturing. The macroeconomic news releases considered here are based largely on the Bureau of Labor Statistics classifications of major economic indicators. Thus, they are selected because of their anticipated importance. Furthermore, in addition to the coverage given these announcements in Business Week for the U.S. market, the published literature also partially, or in some instances wholly, features these announcements.4

The study contributes to the body of literature on the impact of macroeconomic news releases on prices of financial assets in two main ways. First, it investigates which of the macroeconomic news announcements are the most important in equity valuation. While Pearce and Roley (1985) and McQueen and Roley (1993) document the relationship between stock prices and news releases, this paper investigates the relative importance of different scheduled macroeconomic news releases in stock valuation. Relative importance can be examined by measuring implied volatility changes around scheduled news releases. Moreover, the study contributes to the literature focusing on the impact of macroeconomic news releases on the pricing process of other financial assets than stocks, whose pricing processes are different from those of equity. For example, Ederington and Lee (1993, 1996) report results using data from interest rate and foreign exchange rate markets, Fleming and Remolona (1999) using data from the U.S. treasury market, and Christie-David, Chaudhry, and Koch (2000) using data from gold and silver markets.

Second, the study extends previous studies such as Ederington and Lee (1993, 1996), Fleming and Remolona (1999), Christie-David, Chaudhry, and Koch (2000), and Nikkinen and Sahlstrom (2001) by investigating the effect of release day of the month on the importance of macroeconomic news. Macroeconomic reports contain information on the last month's or quarter's economic activity. Furthermore, the information content of some macroeconomic indicators is likely to be similar. Thus, inferences could be drawn regarding coming news releases as a result of previous releases during the same month. For example, inferences regarding the CPI may be made as a result of previous information releases pertaining to NAPM (manufacturing), Employment Report, Producer Price Index, and Import and Export Price Indices. Hence, it may be the case that some crucial elements of the information releases may have already been reflected in stock valuation. As a consequence, the first announcement of the month can be hypothesized to be more valuable relative to subsequent announcements.

The rest of the paper is organised as follows. The following section provides a theoretical underpinning of using implied volatilities to measure the importance of scheduled announcements. In the third section, we present the data used in the analyses in the paper. The fourth section discusses the research methodology. Empirical results are presented in the fifth section, and the sixth concludes the paper.

Measuring the Relative Importance of Scheduled Announcements Using Implied Volatilities

Scheduled macroeconomic announcements provide important information for stock market investors, who use this information to re-assess the valuation of stocks. Since there is a great deal of uncertainty and, thus, disagreement about the content of the coming announcement, actual asset price movement, i.e. realized volatility, tends to be higher than normal on scheduled news announcement days. This phenomenon has been documented by Ederington and Lee (1993); Christie-David and Chaudhry (1999); Fleming and Remolona (1999); and Christie-David, Chaudhry, and Koch (2000) among others. After the announcement, information is incorporated into prices and uncertainty in resolved. Hence, the price adjustment as a result of the information content of scheduled macroeconomic news announcements is reflected in stock market volatility.

To measure the relative importance of macroeconomic announcements, the option pricing theory is applied. Specifically, the evolvement of uncertainty in the stock market is analyzed. Uncertainty on the stock market is measured by using implied volatilities estimated from index option prices. Implied volatility can be interpreted as a market's expectation of the average return volatility over the remaining life of the option contract (Merton 1973). Consequently, the uncertainty around a scheduled news release is reflected in implied volatility. Thus, instead of investigating the investors' reactions on stock markets, this paper examines changes in investors' expectations of uncertainty around macroeconomic news announcements. The purpose is therefore not to analyze the price reaction due to unexpected change in macroeconomic variables (Donders and Vorst 1996; Ederington and Lee 1996).

According to option theory, the option value is a nonlinear function of five factors. These are the underlying stock price, the time to expiration of the option, the exercise price of the option, the risk-free interest rate and the underlying asset price volatility over the remaining life of an option (Black and Scholes 1973). Except for the volatility, all values of the above-mentioned variables are known in advance. Therefore, the current value of an option 0^sub t^ can be expressed as a function of volatility [sigma], 0^sub t^ = [function of]([sigma]). When the market price of the option is also known, implied volatility [sigma].^sub isd^ is given by [sigma]^sub isd^ = [function of](0^sub t^) where [function of]^sup -1^ denotes the inverse function of [function of] The implied volatility estimate is the market's assessment of the future volatility over the remaining life of the option.

The Black-Scholes model assumes that daily stock returns are independently and identically distributed random variables. Therefore, the daily variances are additive and the average implied variance over the remaining life of the option contract can be calculated by summing the individual daily variances and dividing this sum by the number of days until the expiration date (Merton 1973). Based on this, assuming that one scheduled macroeconomic announcement is made during the life of the option, the average implied volatility [sigma]^sub average^ until the expiration day of the option can be expressed as

where [sigma]^sup 2^^sub NAD^ and [sigma]^sup 2^^sub AD^ are the stock return variances on a non-announcement day (NAD) and on the macroeconomic announcement day (NA), respectively, and T is the number of days to expiration date of the option (Donders and Vorst 1996). The theoretical implied volatility can now be described as a function of time as Merton (1973) shows.5 Figure 1 presents the behavior of implied volatility when investors expect that volatility will be constant on non-announcement days and twice as much as non-announcement day's volatility on the announcement day. Implied volatility increases prior to the macroeconomic announcement and then drops immediately after the information is released. The size of this effect can be used to measure how important the different macroeconomic news is considered by stock market investors.

Data Description

Macroeconomic Announcements

The sample of scheduled macroeconomic news releases investigated is largely based on the Bureau of Labor Statistics (BLS) classifications of major economic indicators. Thus, they are selected because of their anticipated importance. Furthermore, in addition to the coverage given these announcements in Business Week for the U.S. market, the published literature also partially, or in some instances wholly, features these announcements. This sample consists of the U.S. macroeconomic news releases covering the period between January 1995 and December 2001. The releases are made in the morning when the stock market is not open. The average dates of the announcements, the issuing authorities of the information releases, and number of announcements contained in the sample are presented in Table 1.

VIX Volatility Index

The measure of uncertainty used in this paper is the implied standard deviation (ISD) calculated from option prices. That is, the daily closing values of the CBOE Market Volatility Index (VIX) are used as a proxy for the market's uncertainty. Since the results concern firms belonging to the S&P100 index, which consists of the largest U.S. firms, the results may not be fully generalized, for example, to small firms. Blair, Poon, and Taylor (2001) indicate that using VIX is advantageous given that it mitigates most of the problems associated with the measurement of implied volatilities. For instance, VIX is likely to minimize the biases caused by the transaction costs (Day and Lewis 1988). In addition, given that the VIX index represents at-the-money implied volatility, it minimizes the magnitude of bias induced by time-varying stochastic volatilities (Day and Lewis 1988).

VIX is calculated as the average of eight S&P 100 index put and call option implied volatilities. The volatilities are calculated using the cash-dividend adjusted binomial method because the OEX options are American type options. VIX is constructed from the option series with the fewest, but has at least eight, calendar days to expiration and from the series of the following contract month. Four pairs of put and call options, whose strike prices are either just above or below the index price, are weighted according to the following three steps:6 First, the implied volatility of put and call options whose strike prices are just above the index price is averaged. Correspondingly, an average implied volatility is calculated using put and call options whose strike prices are just below the index price. Averaging the call and put implied volatilities alleviates possible bias due to staleness in the observed index level. Second, at-the-money implied volatilities are interpolated using these average implied volatilities. Third, given that two sets of options (one set with shorter maturity options and one set with longer maturity options) are used, there are two interpolated at-the-money implied volatilities with different maturities. These volatilities are weighted to create a single volatility that always has 30 calendar (22 trading) days to expiration.

Table 2 provides descriptive statistics of implied volatilities for the sample covering the period between January 1995 and December 2001. The statistics show that the implied volatility varies considerably. For example, minimum value is 10.49 percent, while the maximum is 49.04. Figure 2 presents a visual representation of the behavior of VIX during the sample period. It can be seen that the implied volatility level has increased during the sample period. Moreover, there exist some periods with high implied volatility.

Research Methodology

Two different kinds of methodologies have previously been applied when investigating the effects of scheduled news on implied volatility. Donders and Vorst (1996) investigate the impact of earnings announcements on implied volatility, while Ederington and Lee (1996) examine the impact of macroeconomic news announcements. While earnings announcements occur on average every three months, there are several macroeconomic news announcements every month. Therefore, it is not possible to calculate a mean or median value for implied volatility not containing macroeconomic news announcements as required in the methodology of Donders and Vorst (1996). Consequently, this study uses the methodology of Ederington and Lee (1996).

Implied volatility is hypothesized to gradually increase prior to the scheduled macroeconomic news announcements as presented in section 2. On the announcement day implied volatility is hypothesized to revert to its normal level. The following regression model is estimated to examine the impact of the macroeconomic news releases on implied volatility:

where [sigma]^sub t^ is the implied volatility value at day t and index i refers to a particular macroeconomic news announcement of the same order as in Table 1. The dummy D^sup MacroNews^^sub i,t^ has a value of one on a particular macroeconomic news announcement day and otherwise zero. D^sup FOMC^^sub t^ is a similarly defined dummy variable for the Federal Open Market Committee (FOMC) announcements. It is included in the model since, as a scheduled announcement followed by market participants, it may have an effect on implied volatility as shown by Nikkinen and Sahlstrom (2003). Based on the theory presented in the second section, it is hypothesized that [alpha] > 0, since on days with no scheduled news releases implied volatility is expected to increase.

Godfrey's test indicates that significant first-order autocorrelation is present; thus, an AR(1) term is added to the model. Moreover, the Lagrange Multiplier (LM) test for the ARCH effect indicates conditional heteroskedasticity in error terms. Therefore, GARCH(1,1) is fitted. The coefficients of the GARCH equation are statistically significant, and the LM test shows that the specification is adequate. Based on the Weisberg (1985) test, eight outliers are identified at the 5 percent significance level. The results are reported with and without outliers.

Two non-parametric test procedures are used to gain further understanding of the daily behavior of implied volatility around macroeconomic news releases. These are the standard sign test and the Wilcoxon sign-rank test. For the sign test, under the null hypothesis, the positive and negative changes in implied volatility are equally likely. The Wilcoxon test is used to test the null hypothesis that the sum of positive ranks equals the sum of negative ranks. The changes are again measured as logarithmic differences.

To investigate the effect of release day of the month on the importance of macroeconomic news the following model is estimated:

where the dummy variable, D^sup MacroNews^^sub t^ is one if some scheduled macroeconomic news is released and otherwise zero. T^sub t^ is a square root of the number of calendar days from the beginning of the month to the release day. The results are not sensitive to the functional form of T^sub t^. The assumptions of the regression model are assessed and violations dealt with as before.

Results

The regression results of Equation (2) are reported in Table 3. The table includes results for the raw data as well as data with outliers removed. The results of the regression suggest a differing impact of the announcements on stock valuation. That is, some macroeconomic news releases are more significant in relation to others in stock valuation. Of the 11 macroeconomic announcements examined, only five are found to exert a significant influence on stock valuation. These five macroeconomic news releases are found to be of importance in stock valuation, all having significant negative slope coefficients. Similarly, the coefficient of FOMC is negative and significant. The intercept term is significantly positive indicating that, on average, implied volatility increases during no news days as hypothesized.

The significance level indicates the relative importance of the associated variable. The results from the regression analysis indicate that the Employment Report and NAPM (manufacturing) exert the greatest impact on stock valuation (significance at 1 percent level). This result holds for both the raw data and the dataset with outliers removed. Furthermore, the results suggest that the Producer Price Index, the Import and Export Price Indices, and the Employment Cost Index have a significant effect on implied volatility at the 5 percent significance level even though the coefficients of these announcements are significant only at the 10 percent level in the case of the raw dataset. Moreover, the Productivity and Costs Index has a significant effect on the raw data, but the removal of the outliers removes the effect.

As noted above and as can be seen from Table 3, the NAPM (non-manufacturing), Retail Sales, CPI, Real Earnings, and GDP do not appear to exert any measurable influence on stock valuation. That is, the information content of these announcements does not cause drops in implied volatility after release. This may be surprising given that the information content of these variables reveals important details of economic activities. However, it may be the case that the preceding news releases contain information revealed by the macroeconomic news releases with insignificant coefficients. That is, informed inferences could be drawn regarding these news releases as a result of previous releases during the same month. For example, inferences regarding the CPI may be made as a result of previous information releases pertaining to NAPM (manufacturing), Employment Report, Producer Price Index, and Import and Export Price indices. Hence, it may be the case that some crucial elements of the information releases may have already been reflected in stock valuation.

Non-parametric tests are also conducted to establish the robustness of the findings presented in Table 4. In these tests, the days with more than one announcement are omitted. This resulted in a lower number of observations and, as a consequence, in a lower test power. The results of these tests are shown in Table 4. The results of the tests are in conformity with the results presented in Table 3. The results confirm that NAPM (manufacturing) and the Employment Report exert the strongest impact on implied volatility. The number of positive changes, on the announcement day, for the two variables is 29.7 and 24 percent, respectively. The test signs indicate that both proportions are statistically different from 50 percent (p-values are 0.000). Moreover, the implied volatility decreases, on average, by 2.4 and 4.0 percent, respectively, for NAPM (manufacturing) and Employment Reports. These changes are statistically significant according to the Wilcoxon sign-rank test. Turning attention to an announcement that has the least relative impact on implied volatilities, among the six announcements that have significant coefficients in the previous regression analysis, it could be seen from Table 4 that the number of positive changes, on the announcement day, relating to employment cost index is 44.4 percent. The p-values associated with this, according to the sign test, are 0.073. Generally, the results of the non-parametric tests are well in line with the regression results.

It was hypothesized above that preceding news releases could contain information revealed by the macroeconomic announcements with insignificant coefficients. Hence, the effect of time on the impact of macroeconomic news releases on implied volatility is examined by estimating Equation (3). The results presented in Table 5 confirm the conjecture. Time is found to have a moderating impact on the relationship between macroeconomic announcements and implied volatility since the coefficient of the announcement time is significantly positive. In addition, the intercept term is significantly positive and the coefficient of the release day significantly negative, both of which are in accordance with expectations.

Conclusion

This paper investigates the relative importance of macroeconomic news releases for stock valuation. The study focuses on 11 macroeconomic announcements, which are selected based on the previous literature and the Bureau of Labor Statistics classifications of major economic indicators. Thus, they are selected because of their anticipated importance. Data from the U.S market covering the period between January 1995 and December 2001 are used in the analyses.

This paper finds that five out of the 11 announcements have significant influence on stock valuation. These are the Employment Report, NAPM (manufacturing), Producer Price Index, Import and Export Price Indices, and Employment Cost Index. Of the five announcements, the Employment Report and NAPM (manufacturing) have the greatest impact on stock valuation.

The macroeconomic reports measure the previous month's or quarter's economic activity. Thus, inferences could be drawn regarding upcoming news releases as a result of previous releases during the same month. As a consequence, the first announcements of the month can be hypothesized to be more valuable than the later announcements. This hypothesis is confirmed by the results of the study.