Forecasting Oil & Gas Employment for the State of Wyoming

Forecasting Oil & Gas Employment for the State of Wyoming

by: David Bullard, Senior Economist

"An increase of one operating rig in Wyoming is associated with an increase in oil & gas employment of almost 20 (19.417) individuals over the previous month."

Employment in the oil & gas industry is very important to Wyoming's economy. In 2001, this industry accounted for an average of 11,800 jobs (4.8% of total nonfarm employment). In contrast, the oil & gas industry only accounted for 0.3 percent of U.S. nonfarm employment. In Wyoming, the 2001 average annual wage for oil & gas jobs was $48,000, well above the statewide average wage of $28,000. Using data from 1992¹ to the present, this article presents a model which can be used to predict monthly employment in Wyoming's oil & gas industry. We expect employment to rise with energy prices and drilling activity within Wyoming.

Employment in Wyoming's oil & gas industry varies greatly from month to month and year to year. It rose from only 6,700 jobs in April 1996 to 12,700 during late 2001. Understanding and predicting oil & gas employment is important for understanding state tax revenues² as well as Wyoming's economy as a whole.

For purposes of this analysis, the oil & gas industry is defined as firms in Standard Industrial Classification (SIC) 13. It includes firms engaged in crude petroleum and natural gas production, drilling oil and gas wells, oil and gas exploration, and all other oil and gas field services. Other employment related to oil and gas activity is found in Manufacturing (oil refineries); Transportation, Communications, & Public Utilities (oil and gas pipelines); and Wholesale Trade (oilfield equipment sales and service). However, those industries are outside the scope of the present analysis.

Research & Planning (R&P) publishes monthly employment data for the oil & gas industry. Employment data are estimated from the establishment survey³ and contain sampling error. However, through the benchmarking process, employment data are revised based on a near-universe count using administrative data from Unemployment Insurance (UI) tax reports.⁴

Data Model

According to a regression model,⁵ the level of oil & gas employment in Wyoming is strongly and positively associated with national energy prices, the reported rig count in the state, and additive seasonal factors. Monthly data on energy prices and drilling activity are readily available from various sources to use in our analysis.

Our price variable is the Consumer Price Index for all Urban Consumers (CPI-U) index for utility natural gas service. This index is published monthly by the Bureau of Labor Statistics (BLS) and is available at a one-month time lag (i.e., May data are available in mid-June).⁶ Price is measured as an index where 1982-1984 prices=100. During the period in question (1992-present), the index ranged in value from 97.1 in March 1992 to 186.9 in January 2001, with a mean value of 116.24. Figure 1 shows the behavior of this price index and oil & gas employment from 1992-2001. Notice the large run-up in prices in late 2000 and early 2001, which returned to lower levels by October 2001.⁷

For the reported rig count, we use the Baker Hughes North American Rotary Rig Count data for Wyoming.⁸ Data are available on a weekly basis, with about a one-week lag. We use the monthly averages for our model. From 1992-2001, the rig count ranged from 8 in March 1993 to 65 in September 2001, with a mean value of 36.4. This variable is measured as the number of operating rigs in the State. At its peak, 65 different drilling rigs were operating within Wyoming. Figure 2 shows the rig count and oil & gas employment from 1992-2001. It illustrates how oil & gas employment rises and falls with each large change in the rig count.

We expect that the coefficients⁹ for both the gas price index and rig count will be positive. Increases in energy prices tend to spur exploration and production which increases employment. Similarly, expanded drilling activity (i.e., the number of rigs operating) also increases employment.

It may be necessary to justify the use of both price and rig count, because they are correlated with each other (r=.559). Regardless of how high national energy prices go, employment in Wyoming could remain low if proven reserves¹⁰ within the State were low or if environmental regulations made exploration and drilling difficult. Similarly, not all oil & gas activity within the state is captured by the Baker Hughes rig count. A significant share of coalbed methane drilling uses rigs too small to be counted in the Baker Hughes series.¹¹

Wyoming's severe winter weather tends to hold employment down. Employment tends to peak during the late summer when weather is not a factor affecting outdoor work. Therefore, we expect the seasonal factors to be positive (the month of January is the base case).

Results

We use Ordinary Least Squares (OLS) regression to estimate the effect of changes in prices and rig count on oil & gas employment in Wyoming. Linear regression is a commonly used statistical technique in which researchers are able to estimate the effect of one independent variable on a dependent variable, while holding the other independent variables constant. Table 1 summarizes results of the regression model.¹² The adjusted R2 is 0.676, suggesting that the model explains over two-thirds of the over-the-month change in oil & gas employment.

The results of the model indicate that, holding energy prices and seasonal factors constant, an increase of one operating rig in Wyoming is associated with an increase in oil & gas employment of almost 20 (19.417) individuals over the previous month. Similarly, holding the change in rig count and seasonal factors constant, a one-point increase in the CPI-U for utility natural gas service will increase Wyoming oil & gas employment by approximately 4 (3.811) individuals over the previous month. The seasonal factors show that holding prices and the change in rig count constant, the change in employment is higher in the summer months and lower during the winter. January is the lowest month and the peak is in June, when the change in employment is 602 (601.704) individuals higher than in January.

Figure 3 shows the actual change in oil & gas employment compared to the predicted changes for January 2000 through May 2002. Readers should remember that the actual series will be revised again in March 2003 through the benchmarking process.

Conclusion

Oil & gas employment is an important variable because of its significance to Wyoming's economy and tax revenues. This model uses national energy prices, a rig count for the state, and seasonal factors to predict employment. Results using actual data from 1992 through 2001 show that the model has significant predictive power.

¹We start the model with 1992 because that is the first year for which Baker Hughes published a monthly rig count by state.

²Oil & gas severance taxes account for roughly 10 percent of general fund revenues. See the Consensus Revenue Estimating Group (CREG) monthly report at <http://eadiv.state.wy.us/creg/cregbrief.pdf>.

³The establishment survey, also known as the Current Employment Statistics (CES) program, is a joint federal-state cooperative survey which asks employers to report the number of employees on their payrolls each month. These data are used to estimate employment by industry for the nation, all 50 states, and a large number of metropolitan areas.

⁴The revised employment figures are released in March of the following year. Thus, while January through May 2002 employment data are used to test the model, we expect that these numbers will be revised (in March 2003).

⁵We use statistical models to explain and/or predict economic phenomena. In this case, we want to explain and predict employment in the oil & gas industry. Thus, oil & gas employment is the dependent variable, meaning it is dependent upon other factors. The factors that we use to explain employment (natural gas price and rig count) are called independent variables. In our model, natural gas price and rig count are the inputs and an estimate of oil & gas employment is the output.

⁶U.S. Department of Labor, Bureau of Labor Statistics, “Consumer Price Index for all Urban Consumer (CPI-U) utility natural gas service,” Consumer Price Indexes, n.d., <http://stats.bls.gov/cpi/home.htm> (July 26, 2002).

⁷Readers interested in producing a longer-term forecast for employment might use price data available in the Annual Energy Outlook 2002 with Projections to 2020, available from the Energy Information Administration at <htp://www.eia.doe.gov/oiaf/aeo/index.html>.

⁸Baker Hughes, North American Rotary Rig Count, Rig Counts, n.d., <http://www.bakerhughes.com/investor/rig/rig_na.htm> (July 26, 2002).

⁹The unstandardized regression coefficient (b) can be interpreted as increases or decreases in employment (depending on a positive or negative sign) for a one-unit increase in the variable of interest.

¹⁰Proven reserves refer to oil or gas that is still in the ground and has not been extracted. In contrast, oil which has been extracted and is being stored in tanks is known as inventory.

¹¹"The Baker Hughes Rotary Rig count includes only those rigs that are significant consumers of oilfield services and supplies and does not include cable tool rigs, very small truck mounted rigs or rigs that can operate without a permit." Baker Hughes, “North American Rotary Rig Counts,” Rig Counts, n.d., <http://www.bakerhughes.com/investor/rig/rig_na.htm> (July 26, 2002). Often, coalbed methane is drilled using truck mounted rigs.

¹²The Durbin-Watson statistic indicated that positive serial correlation was present in the original model (DW=.222), so we took first differences of the rig count and the employment series to arrive at an adjusted model. Because of the transformation to first differences, the adjusted model predicts the monthly change in employment based on the price level, the change in rig count, and the seasonal factors. The Durbin-Watson statistic for the adjusted model (2.046) is within the 1 percent range to accept the null hypothesis that no serial correlation is present.

Table of Contents | Labor Market Information | Wyoming Job Network | Send Us Mail

These pages designed by Julie Barnish.
Last modified on by Susan Murray.