Skip to content
Quick Start for:
A Revised Texas Index of Leading Economic
Indicators for the New Millennium

The Texas State Comptroller's office revised its Texas index of leading economic indicators to better predict changes in Texas economic activity in the new millennium.

Comptroller economists noticed an increasing tendency of the old index, last revised in 1990, to underpredict future nonfarm employment growth.1 They realized that a revised indicators index was needed to better track the relatively strong Texas economic growth of the 1990s, while still reflecting the state's boom and bust economy of the 1980s.

The past year was dedicated to developing what comptroller economists believe to be a vastly improved leading economic indicators index (see Figure 2). The revisions will create a more accurate, current indicator of overall state economic activity. They were accomplished through a two-step process:

  1. The weight of each of the 10 index components was determined based on its accuracy in predicting Texas economic activity, with a six-month lag, in the 1980s and the 1990s.

  2. The simple linear equation was re-estimated to reflect the growth of statewide nonfarm employment, which includes the significant increase in high-tech related jobs and the dramatic drop in oil and gas jobs.
New Component Weights
Two distinct methods were used to determine the new component weights in the updated leading economic indicators index.

Method 1: Three Comptroller economists were shown graphs of the year-over-year growth in each of the 10 indicators, with a six-month lag, compared to the growth in nonfarm employment from 1981 through 1999. The economists were asked to rate each of the indicators on a scale of 0 to 10 for accuracy in predicting state economic growth. The average was taken of the three economist's scores and converted to a relative weight, based on its share of the total average score awarded to all 10 economic indicators.

Method 2: At the same time, another Comptroller economist independently experimented with ways to re-adjust the component weights to keep the index more on track in recent years. Surprisingly, the component weights developed by this independent analysis were similar to those developed by the three-economist scoring (see Method 1, above).

The weights from the two separate analyses were then averaged to determine the final component weights.

While updating the index, several other components used in the U.S. leading economic indicators index were considered. These include the spread between long-term and short term Treasury bond interest rates, the M2 money supply, new orders received by non-defense capital goods manufacturers, and vendor performance (measuring the percentage of manufacturers experiencing slower deliveries). However, none of these alternative national-level indicators proved effective at predicting the Texas economy, while the 10 older Texas-specific indicators remained accurate predictors.

Therefore, we re-weighted the 10 predictive components used by the earlier index to create this revised index, which is much more accurate than either the previous version or the other versions tested.

For example, average weekly manufacturing hours accounted for 48 percent of the old index, but because Texas manufacturers are currently more likely to change the size of their workforce before adjusting worker hours, this share has declined to 5 percent in the new index (see Table 9). In place of manufacturing hours, initial claims for unemployment insurance, the level of help-wanted advertising and consumer confidence now make up nearly half of the index.

Table 9

Also, indicators such as new business incorporations, residential building permits, and the Comptroller's stock index--all of which reflect a growing high-tech sector--now make up nearly a quarter of the index. Finally, because of the important role that the national economy plays in determining Texas' economic performance, the U.S. leading economic indicator index accounts for approximately 15 percent of the revised Texas index.

Re-estimated Translation Equation
After the component weights are estimated, a simple linear equation is used to translate the weighted index into a prediction of monthly Texas nonfarm employment growth. The following linear regression is estimated:

      B * MOVAVG(3, PCHYA (WINDEX[-6]))

where PCHYA (XTNEM) is the year-over-year percentage change in nonfarm employment and MOVAVG(3, PCHYA (WINDEX[-6])) is the three-month moving average of the year-over-year growth in the weighted leading economic indicators, lagged six months.

A regression estimated on monthly nonfarm employment and the weighted leading indicators from 1981 through 1999 yielded an estimated CONSTANT term of 2.049 and an estimated B coefficient of 0.201, with both parameter estimates highly significant at better than the 0.05 percent level. Overall, the estimated equation explains 83 percent of the variance in the year-over-year growth in Texas nonfarm employment during this period.

These parameter estimates have two implications. First, over the long term, Texas nonfarm employment grows at approximately 2 percent per year, regardless of external economic conditions. Second, after accounting for this baseline 2 percent rate of growth, statewide nonfarm employment increases at approximately 20 percent the growth rate of the newly re-weighted indicators.

The Revised Monthly Leading Indicators Model
Now with the parameters estimated, we can assemble the leading indicators model to produce monthly estimates of the total index.

Since the index is initially calculated on a percentage change basis, the first step of this process is to convert each of the 10 index components into monthly (symmetric) percentage changes. For example, the monthly percentage change (RHWINX) of the help-wanted index (HWINX) is calculated as:

(1) RHWINX = 200 * (HWINX-HWINX[-1]) / (HWINX+HWINX[-1])

This calculation is subsequently repeated for each of the 10 index components.2

Following the convention used by The Conference Board in the construction of the U.S. economic indicators, a symmetric percentage change is used to treat increases and decreases in the indicator equally. With a symmetric percentage change, a one-percent increase followed by a one-percent decline in an indicator leaves the variable at its original level. This is not true with the conventional percentage-change formula, where the same percentage increase and decline in the variable would leave it slightly below its original level.3

After the monthly percentage change for each of the indicator components is calculated, a weighted average percentage change of all 10 index components, based on the weights determined earlier, is computed.

(2) RPLEADN = (0.111*RXEINITCL) + (0.261*RHWINX) +
      (0.108*RXJCCWSC) + (0.048*RXMFWH) + (0.049*RXRETNA83) +
      (0.145*RUSLEAD) + (0.049*RXSTOCKN83) + (0.053*RXHPERMIT) +
      (0.114*RXINCORP) + (0.062*RXOILPR83)

Next, this weighted average is translated into a prediction of future employment growth (SRPLEA) using the constant and slope parameters estimated above. Because the leading indicator index is calculated on a month-to-month rather than a year-over-year basis, however, the constant term that is used in this equation (0.19) is approximately one-twelfth of the year-over-year estimate of 2.049 calculated above.4

(3) SRPLEADN = 0.19 + 0.201* RPLEADN

Finally, the month-to-month percentage change in the final leading indicator index is converted to an actual value by solving the symmetric percentage change formula for SRPLEADN = 200 * ((PLEADN-PLEADN[-1]) / (PLEADN+ PLEADN[-1])) for PLEADN.

(4) PLEADN = PLEADN[-1] * (200+SRPLEADN) / (200-SRPLEADN)

and then updating the monthly PLEADN estimate on a recursive basis.

1 See "Comptroller's Index Heralds Job Changes," Fiscal Notes, April 1990, p. 12.

2 For this calculation, each of the 10 component series is seasonally adjusted at an annual rate. In addition, to ensure comparability with a series where an increase in an indicator projects a slowdown in economic growth (initial claims for unemployment insurance), the index is multiplied by -1.

3 See "Calculating the Composite Indicators" in The Conference Board on the Web (

4 Actually, the twelfth root of an average annual growth rate of 2.049 percent translates to a monthly growth rate of 0.17 percent. For prediction purposes, however, this monthly estimate was increased to 0.19 because it fits the rapid nonfarm employment growth of the second half of the 1990s more accurately.