According to Air Force Policy Directive 32-30, paragraph 1.1, "The Air Force must sustain the capability to disarm unexploded ordnance delivered or place by enemy forces. In addition, the Air Force must be able to "render safe" US ordnance made dangerous by accident or other circumstance. The Air Force is also obligated to use its special expertise to assist Federal and civil authorities when called upon in times such as dealing with terrorist or other criminal acts, accidents and found explosive items." To meet this need the Air Force has what it calls Explosives Ordnance Disposal (EOD) teams. This study does not address the entire EOD career field but focuses only on the portion of the career field that deals with explosive items found on military practice bombing ranges. The term generally used to describe the process of picking up ordnance on practice ranges is "Range Clearance".

To clear the practice bombing ranges, it takes experienced EOD technicians, specialized equipment and lots of time. The clearance must be scheduled in advance with the range control operators, local flying organizations, local health care providers (to obtain medical support) and the EOD unit itself. In recent years, the drastic military cutbacks have had a significant impact on military occupations. Some occupations have completely disappeared merging with other, similar, occupations or perhaps they were determined to no longer be necessary and totally eliminated. Others have been reduced to a fraction of their previous strength. The Explosive Ordnance Disposal (EOD) career field, Air Force Specialty Code 3E8X1, is such a career field. The manning authorization was reduced 8.5% from 1996 to 1998.

Just how many people does it take to properly and safely clear a range? That is the question that this study addresses. Several major factors were identified as affecting the personnel requirements for range clearance such as the type of ordnance, climate, and terrain. Thus, the purpose of this study was to gather historical data, analyze requirements, and provide conclusive factors to accurately determine civilian and military manpower requirements for Explosive Ordnance Disposal (EOD) personnel to perform explosives decontamination and thermal treatment on Air Force Weapons and Training Ranges.

METHODOLOGY

The source of the historical data used for the analysis was taken primarily from the AF Form 3578, Explosive Ordnance Disposal Report. The historical data was not always complete and no acceptable methodology existed for back-filling missing data. Thus, the data, which was used for the analysis, represents a sample of all the range clearances performed during the years 1992 through 1995. In addition, the observations, which were included in each of the analyses, are a result of the data available to support each respective analysis and model specification.

The first stage of the analysis was to identify or develop a methodology for determining the amount of manpower necessary for range clearance. Through discussion with subject matter experts and analysis of the Explosive Ordnance Disposal Reports for 1992 through 1995, the approach selected for modeling the range clearance was as a production function. Range clearance produces three primary outputs: tonnage of ordnance collected, acres cleared, and quantity of nomenclature collected. The type of ordnance, which is listed in the Explosive Ordnance Disposal Reports, is assumed to affect the number of man hours to complete its disposal since it may require additional detonation. Other activities are performed during the range clearance such as training and equipment maintenance and cleaning. These additional activities are not directly captured through the production function approach but will be assumed to be captured by the intercept term for the estimated production function, invariant to the level of clearing, i.e., tonnage or acres cleared.

The production function will be modeled as a multi-product production function in which the number of man hours is the predominant determinant but other factors must for considered such as terrain and weather. Thus, the model can be generically specified as

where

T is the tons of scrap

A is he acreage cleared

QN is the quantity of ordnance which required special handling

mh is the man hours used in the range clearance

season_I represents the I^th season of the year

terrain_j represents the j^th type of terrain, e.g., mountainous, dessert, swampy, etc.

year_t represents the t^th year, 1992 through 1995

One aspect of a multi-product production function should be noted. The number of man hours used in the range clearance directly affects the magnitude of tonnage, acreage and quantity of ordnance and the respective magnitudes of tonnage, acreage and quantity of ordnance affect each other. For example, assume a range clearance which involves a fixed number of acres and a fixed number of man hours. Thus, the amount of tonnage and ordnance cleared will experience tradeoffs between each other based on the magnitudes of each. In order to increase the tonnage in the range clearance, the range clearance crew can not be involved in a large quantity of ordnance requiring special handling. This constraint can be minimized by providing more man hours which would allow increases in both tonnage and quantity of specially handled ordnance. Given that tradeoffs can occur between all three outputs, the best model for such a production function is a simultaneous system of seemingly unrelated equations. The specification for such a system would be

where an interaction is assumed to occur, causing the quantitative values for T, A, and QN to be related to each other. This can be represented by

which implies that the quantitative value for one of the outputs affects the quantitative value for the other two.

In the estimation of a simultaneous system the interaction of the dependent variables is reflected in the off-diagonal elements of the variance-covariance matrix of the system. If these off-diagonal elements are non-zero (statistically different from zero), then the equations of the system are related, and the simultaneous system estimation is the proper specification versus three stand-alone singel equation specifications. The estimator for the simultaneous system of equations will be Zellner’s seeming unrelated regression (Zellner, 1962, Zellner and Huang, 1962, and Zellner, 1963) and use the asymptotically efficient, feasible generalized least-squares algorithm (Greene, 1993). Two mathematical specifications for the production function were used: linear and log-linear. The log-linear specification is generally referred to as a Cobb-Douglas production function (Henderson and Quandt, 1958). Both the linear and Cobb-Douglas specifications were estimated, but only the linear results are presented below.

In addition to the estimation of the simultaneous equation model, a reduced form model was also estimated for comparison to the results of the estimated simultaneous model. The reduced form was specified as

The reduced form equation specifies acreage cleared as a function of tonnage and quantity of ordnance (explanatory variables). The reduced form equation was estimated using ordinary least-squares regression in linear and log-linear formulations. Once again, only the linear results are presented.

Tables 1 presents the results concerning man hour requirements of the seemingly unrelated regressions (SURE) using the linear specification for the production function. The man hours used to perform the range clearances is statistically significant in each of the three equations at the 99% level of confidence. The R-square statistics for the three system equations were 0.322, 0.494, and 0.364 for tonnage, acres cleared and quantity of specific ordnance, respectively, each of which was statistically significant at the 99% level of confidence. The test for significance of the off-diagonal elements of the variance-covariance matrix is statistically significant at the 99% level of confidence, indicating that there are statistically significant covariance (the equations are not independent) between the three dependent variables, tonnage, acres cleared and quantity of specific ordnance.

Table 2 presents the results of the reduced form estimation. The coefficient for man hours is statistically significant a the 99% level of confidence. The value of the coefficient for man hours is not statistically different from the value provided by the simultaneous equation estimation in Table 1, 9.4174 versus 9.5215 (t-test). This substantially validates the model specification of the simultaneous system. The R-square statistic for the reduced form equation was 0.422, which was statistically significant at the 99% level of confidence.

A survey was performed to gather information about the range clearance, which occurred at several bases. Seventeen bases responded to the survey. The largest number of surveys returned was from Nellis (63) and Eglin (15), which comprises over 69% of the sample. Thus, most results of the survey will be supplied by base to minimize potential bias from Nellis and Eglin. The survey identified three primary types of range clearance: 50 use day, annual or partial five year, and five year clearance. The annual or partial five year range clearance was the most often type of clearance performed across bases followed by the 50 use day range clearance.

Table 3 presents the acres per man hour from the AF Form 3578s by year by type of clearance and the survey results. The 50 day use clearances exhibit a definite upward trend over the 1992 to 1995 time period. Annual and 5 year clearances exhibit more up and down fluctuations over the 1992 to 1995 time period. With the exception of 1994, 5 year clearances exhibit substantially higher acres per man hour than annual clearances, and 5 year and annual clearances are definitely higher in acres per man hour than the 50 day use. The acres per man hour for the three types of range clearances across the 1992 to 1995 time period are statistically different from each other at the 99% level of confidence. The annual and 5 year range clearances are statistically different from each other for three of the four years (exception, 1994) at the 95% level of confidence. The survey based mean values of acres per man hour between annual and 5 year range clearances are not statistically different.

The 8.8234 acres per man hour, which represents the mean value across the three types of range clearances from the survey data, is similar to the 9.5215 acres per man hour coefficient from the SURE regression equation for acres cleared (Table 1). SURE coefficient and the survey mean are more comparable than the 4.8164 mean value from the AF Form 3578. Thus, the multivariate analysis performed using the AF Form 3578 data provides an estimate of acres per man hour more comparable to the survey generated mean value than the actual mean for the AF Form 3578 data. The survey data was collected based on "typical" years, while the multivariate analysis attempts to obtain an acres per man hour value (coefficient) which accounts for influences which would cause variations from the "typical" year. Thus, factors such as type of clearance, weather, terrain, and type of ordnance can affect in variations in the acres per man hour. Table 3 also provides evidence that the man hour requirements for range clearances have slightly increased over time.

This study gathered historical data, collected survey information, analyzed requirements, and provided conclusive factors to identify manpower requirements for Explosive Ordnance Disposal (EOD) personnel to perform range clearances on Air Force Weapons and Training Ranges. The findings presented indicate that the manpower requirements for range clearances vary significantly by type of clearance, acres cleared and type of explosive ordnance cleared. Some inconsistencies were identified between the historical data and survey data, but, in general, the two sets of data provided similar results concerning the usage of manpower to perform range clearances. Slight trends were identified by the historical data which suggest that in increase in manpower has occurred for the given acres and type of clearances performed. Overall, man hours required for range clearances were relatively stable over the time period (1992-1995) for which sufficient objective data was available to analyze relationships between man hours and acres cleared, as well as tons of ordnance cleared.

The survey data found little or no difference between the acres per man hour required between annual and five year clearances. Conversely, the historical data (AF Form 3578) indicated an increase in the acres per man hour from 50 day use clearances to annual clearances and annual clearances to five year clearances. This implies that for a given number of acres, the manpower required decreases from 50 day use to annual to five year clearances.

The comparison of historical data with survey data indicates that EOD personnel are performing similar (or the same) types of clearances with smaller crews in 1998/1999 than during the 1992 to 1995 time period. Thus, the findings based on the survey data and historical data suggest that if smaller crews are being dedicated to similar tasks, then one of the critical factors, which may change to compensate for the increased workload, is a reduction in safety. This does indicate, unequivocally, that all range clearances are being performed at higher levels of risk, but it is consistent with such as conclusion. To alleviate the performance risk, work crews would have to have more time or more personnel to perform range clearances. Across ranges and range clearances, 55% more manpower was allocated to range clearances during the 1992 to 1995 time period than in the 1998/1999 survey. Unless the number of tasks performed or the number of acres cleared has declined, EOD personnel are performing these range clearances under greater risk of injury.

Greene, W.H. 1993. Econometric Analysis. 2^nd edition. New York: Macmillan.

Zellner, A. 1962. An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias. Journal of the American Statistical Society 57: 348-368.

Zellner, A. 1963. Estimators for seemingly unrelated regression equations: Some exact finite sample results. Journal of the American Statistical Society 58: 977-992.

Zellner, A. and Huang, D.S. 1962. Further properties of efficient estimators for seemingly unrelated regression equations. International Economic Review 3: 300-313.

Henderson, J.M. and Quandt, R.E. 1958. Microeconomic Theory: A Mathematical Approach. New York: McGraw-Hill Book Co., Inc.
 

---

Next report 

Back to List of Presentations

Back to IJOA main page

posted November 24, 1999