These pages look best viewed in Netscape!
Copyright 1998 by the Wyoming Department of Employment, Research & Planning


County-Based Industry Projections for Wyoming: An Excerpt from the Upcoming Internet Publication

by: Gregg Detweiler

County-based projections for Wyoming and its counties indicate slight improvements in economic conditions for the near future. Wyoming total nonagricultural monthly employment should increase by nearly two percent on average, while Campbell and Teton Counties are projected to see even larger percent changes (3.0 and 4.3%, respectively--see Table 1). But where do these projections come from and how are they best used?

There are various applications for county-based industry projections. One of the most effective uses of county-based projections is impact studies. When forecasting a time series using only historical data, such as monthly employment data (see Figure 1 and Figure 2), the analyst assumes that all independent variables affecting the employment series are constant. Thus, the difference between the projected and actual values can be used to identify economic activity. This includes business births and deaths, opening or closing of special projects, and the movement of a business from one industry to another. For example, in 1992 many oilfield service companies moved their primary business to the heavy construction industry. Other applications of county-based projections are: performance measures of employment/training programs, the ability to target short-term training for welfare recipients and job training partnership act (JTPA) and Current Employment Statistics (CES).

It is often useful to divide a time series into an historical or estimation period and a validation period. One develops a model on the basis of the observations in the historical period and then tests it to see how well it works in the validation period. When you are not sure which model to choose, this technique is sometimes more efficient than comparing models based on the entire sample. In our case, the historical period consists of covered employment data by major industry from January 1989 to December 1996. The validation period was January 1997 through September 1997. We used the historical period to find a model to fit or predict the data series in the validation period. Only seven percent (20) of the time series that were modeled(1) produced an error more than +/- 2.0 percent.

We are all confronted with decision-making situations in which time is an important variable. For example, an economist is interested in predicting monthly employment levels, a store manager is interested in the effect of time on demand for products, and a marketing manager is interested in the pattern of sales over time. In a sense, everyone who plans for the future by attempting to budget time and resources is concerned with processes that are random over time(2).

Why might someone collect such data? What kinds of questions could he or she be trying to answer? One reason to collect time series data is to try to discover systematic patterns in the series so a mathematical model can be built to explain the past behavior of the series. The discovery of a strong seasonal pattern, for instance, might help explain large fluctuations in the data. Figure 1 shows that each of the four time series demonstrates some degree of seasonality. Comparing the two extremes, Campbell County (least seasonality) and Teton County (most seasonality): nine out of the twelve major industries exhibited the same seasonality and five out of twelve industries displayed the same statistical model. Even though these two time series demonstrate the same seasonality, the effects on Teton County are much greater than Campbell County.

Explaining a variable’s past behavior can be interesting and useful, but often one wants to do more than just evaluate the past. The parameters of the model that explained the past values might also be used to predict whether and how much the next few values will increase or decrease. The ability to make such predictions successfully is very important to any business or scientific field.

The type of modeling used was "Exponential Smoothing". The Exponential Smoothing procedure is best used for short-term forecasting, or what are known as "one-period-ahead" forecasts. When you choose the right values for its parameters it extracts a lot of useful information from the most recent observation, somewhat less from the next-most-recent, and so on, usually producing a good forecast for the near-term. As it moves into the future, however, it quickly runs out of the recent information on which it thrives. However, that is when the analyst’s knowledge about the time series plays an important part in how to treat each variable.

One of the reasons Exponential Smoothing was used to forecast each major industry one year in advance is that it is difficult to utilize only past data to predict future economic events unless the analyst has other pertinent knowledge. Also, within the next year, changes in Federal legislation will cause some businesses and employers to be reclassified in a different industry, therefore, causing fluctuations in the employment levels throughout Wyoming’s major industries. Currently, we use a four-digit Standard Industrial Code (SIC); however, that code will change to six-digit North American Industrial Classification System code (NAICS), which will classify employers in the United States, Canada and Mexico into comparable industries.

This article included only four counties (Campbell, Laramie, Natrona and Teton); projections for the remaining counties will soon be available on the Internet via http://onestop.state.wy.us/lmi/rphome.htm. Industry projections for each county will be revised and then forecasted out another three to six months at the end of every calendar quarter, routinely adding new information to the web site.


Gregg Detweiler is a Principal Statistician, specializing in Current Employment Statistics (CES).

1 Statistics for this article were produced using SPSS software.

2 This situation usually occurs in problems where we are attempting to estimate the expected value of a random process or to predict a new value at a future point in time. A time series is a set of observations obtained by measuring a single variable regularly over a period of time. In a series of employment data, for example, the observations might represent monthly employment levels for several months. A series showing housing starts might consist of weekly housing permits taken over a few years. What each of these examples has in common is that some variables were observed at regular, known intervals over a certain length of time. Thus, the form of the data for a typical time series is a single sequence of observations representing measurements taken at regular intervals.


 
Button Bar 4
Go Back To:
Table of Contents
Labor Market Information
Employment Resources
Send Us Mail



These pages designed by Gayle C. Edlin.
Last modified on November 28, 2000 by Valerie A. Davis.