Boynton et al. Journal of Energy Finance and Development 4 1999 1–27 3
We also measure the Spearman rank correlations between two operating profit ratios return on oil and gas revenues and return on oil and gas assets and each of the
four finding costs ratios. We expect a negative association between rank on operating profit ratios and rank on finding costs ratios, that is, that more profitable firms have
lower finding costs. We are particularly interested in determining if the association between the operating profit ratios and the finding costs ratios is stronger or weaker
for any of the four finding costs ratios in a statistically significant sense. We therefore supplement the rank correlation analysis with rank regression of the operating profit
ratios on rank on the finding costs ratios to test for any significant differences in the association for different finding costs ratios with the operating profit ratios.
We also measure the Spearman rank correlations between EFF and each of the two operating profit ratios. We expect a positive association between rank on EFF
and rank on operating profit ratios, that is, more efficient firms in exploration are more profitable overall.
The balance of our paper is organized as follows: In section 2 we describe the selection of our sample of 119 publicly owned oil and gas firms with data for 1988–1992.
In section 2.1 we define the four finding costs ratios that we examine. In section 2.2, we define the two operating profit ratios we use. In section 2.3, we develop our
Cobb-Douglas regression derived index of exploration efficiency EFF including the coefficient estimates of the Cobb-Douglas production model. The estimated EFF for
1989–1992 for each of the 119 firms is presented.
In section 3 we present descriptive statistics for key data items for the 119-firm sample as a whole and for the full cost and successful efforts subsets.
2
In section 3.1 we present Spearman rank correlations between each of the four finding costs ratios
and EFF for the 119-firm sample as a whole and for the full cost and successful efforts subsets. In section 3.2 we present Spearman rank correlations between each of the
four finding costs ratios and the two operating profit ratios for the 119-firm sample as a whole and for the full cost and successful efforts subsets. In section 3.3 we present
Spearman rank correlations between each of the two operating profit ratios and EFF for the 119-firm sample as a whole and for the full cost and successful efforts subsets.
In section 3.4 we present the rank regression results comparing the four finding costs ratios as a set to EFF. In section 3.5 we present the rank regression results comparing
the four finding costs ratios as a set to each of the two operating profit ratios. Section 4 summarizes our findings.
2. Sample selection of 119 publicly owned oil and gas firms with data for 1988–1992
This paper grows out of research performed under a grant EIA Financial Assistance Instrument DE-FG01-92EI23624 from the Energy Information Administration EIA
of the U.S. Department of Energy to the Institute of Petroleum Accounting IPA at the University of North Texas.
3
As part of the grant, EIA furnished IPA with a restricted-use copy of a database prepared by EIA from proprietary databases to
which EIA was a subscriber.
4
SFAS 69 supplemental disclosures data for publicly held firms are part of the annual reports of the individual firms. The data are also available,
4 Boynton et al. Journal of Energy Finance and Development 4 1999 1–27
for a fee, in a computer-accessible proprietary database, from Arthur Andersen LLP, Houston, Texas.
The 119-firm sample for this paper consists of all publicly owned firms listed on the EIA restricted-use database with non-missing values
5
for each of the five years between 1988–1992 for the following SFAS 69 data items: additions and extensions
to proved oil reserves worldwide thousands of barrels, total proved oil reserves worldwide thousands of barrels, total oil production worldwide thousands of bar-
rels, additions and extensions to proved gas reserves worldwide millions of cubic feet, total proved gas reserves worldwide millions of cubic feet, total gas production
worldwide millions of cubic feet, exploration expenditures worldwide thousands of dollars, development expenditures worldwide thousands of dollars, total revenues
from oil and gas operations worldwide thousands of dollars, total oil and gas assets capitalized costs worldwide thousands of dollars, total oil and gas net income
worldwide thousand of dollars, total depreciation, depletion, and amortization DDA worldwide thousands of dollars, and total exploration expense worldwide
thousands of dollars.
2.1. Finding costs ratios A finding costs ratio is calculated by dividing an exploration-related expenditure
for a period by an estimate of the quantity of oil and gas discovered during the same period Gaddis et al., 1992. The financial expenditure and quantity data are
supplemental disclosures required under SFAS 69. Conventionally, and for this paper, oil and gas quantities are aggregated by converting 6,000 cubic feet 6 mcf of gas to
one barrel of oil equivalent 1 BOE, a conversion based on the energy content of oil and gas.
6
Finding costs ratios are a popular but not universally accepted means of evaluating the performance of oil and gas exploration firms.
7
Many firms now disclose their own calculation of finding costs ratios. The 1999 PricewaterhouseCoopers Survey of U.S.
Petroleum Accounting Practices reports that 34 of the 39 responding independent
producers and 5 of the 6 responding major producers calculate finding costs ratios for internal use and that 18 of the 39 responding independent producers and 3 of the
6 responding major producers disclose finding costs ratios externally Pricewaterhouse- Coopers and the University of North Texas Institute of Petroleum Accounting, 1999.
Disagreements exist as to the appropriate exploration-related expenditures to in- clude in finding costs ratios e.g., SFAS 69 exploration expenditures only, the sum of
SFAS 69 exploration and development expenditures, or the sum of SFAS 69 explora- tion and development expenditures and acquisition expenditures, i.e., expenditures
to purchase reserves, the length of the period to use e.g., 1, 3, or 5 years, and the composition of the quantity estimates to be used for example, SFAS 69 additions
and extensions to proved reserves only, the sum of SFAS 69 additions and extensions to proved reserves and revisions to proved reserves [see Clinch and Magliolo, 1992,
for a discussion of the value relevance of revisions to proven reserves], or the sum of SFAS 69 additions and extensions to proved reserves, revisions to proved reserves, and
proved reserves acquired, i.e., purchased. This paper will examine four of the possible
Boynton et al. Journal of Energy Finance and Development 4 1999 1–27 5
variations using two definitions of exploration-related expenditures SFAS 69 explora- tion expenditures only and the sum of SFAS 69 exploration and development expendi-
tures, two definitions of period 1 year and 3 years, and one definition of quantity SFAS 69 additions and extensions to proved reserves only. We believe that these
four will provide insight into the usefulness of finding costs ratios as measures of exploration efficiency and the effect of varying the definition of the finding costs ratio.
The four finding costs ratios that we study are: • FC1 5 exploration expenditures worldwide for year t thousands of dollars
divided by oil and gas additions and extensions to proved reserves worldwide for year t thousands of BOE.
• FC3 5 total exploration expenditures worldwide for three years, years t, t 2 1, and t 2 2 thousands of dollars divided by total oil and gas additions and
extensions to proved reserves worldwide for years t, t 2 1, and t 2 2 thousands of BOE.
• FC1D 5 total of exploration expenditures and development expenditures world- wide for year t thousands of dollars divided by oil and gas additions and
extensions to proved reserves worldwide for year t thousands of BOE. • FC3D 5 total of exploration expenditures and development expenditures world-
wide for three years, years t, t 2 1, t 2 2 thousands of dollars divided by total oil and gas additions and extensions to proved reserves worldwide for years t,
t 2 1, t 2 2 thousands of BOE.
The 3-year finding costs ratios, FC3 and FC3D, use data from both the current year and the prior 2 years in each observation. Using 5 years of data 1988–1992,
we have 3 years for which 3-year ratios may be observed 1990–1992 for each firm, or 357 total observations for each 3-year ratio for the 119-firm sample. To compare
1-year finding costs ratios, FC1 and FC1D, with the 3-year ratios, we also observe the 1-year ratios for 1990–1992 for each firm for a total of 357 observations for each
1-year ratio for the 119 firm sample.
2.2. Operating profit ratios The SFAS 69 supplemental disclosure data also permits the calculation of conven-
tional operating profit ratios for the oil and gas component of a firm. The operating profit ratios that we study are:
• ROS 5 total net income from oil and gas operations worldwide for year t thou- sands of dollars divided by total oil and gas revenues worldwide for year t
thousands of dollars. • ROA 5 total net income from oil and gas operations worldwide for year t
thousands of dollars divided by total oil and gas assets “capitalized costs” worldwide for year t thousands of dollars.
As mentioned above, using 5 years of data 1988–1992, we have 3 years for which 3-year finding costs ratios may be observed 1990–1992 for each firm, or 357 total
observations for each 3-year ratio for the 119-firm sample. For comparability, we also
6 Boynton et al. Journal of Energy Finance and Development 4 1999 1–27
observe the 1-year finding costs ratios for 1990–1992 for each firm for a total of 357 observations for each 1-year ratio. Similarly, for comparability, we observe the 1-year
operating profit ratios, ROS and ROA , for 1990–1992 for each firm for a total of 357 observations for each 1-year ratio.
2.3. Cobb-Douglas regression derived index of exploration efficiency EFF Statistical regression using a Cobb-Douglas production model offers yet another
means of evaluating efficiency in firms in a wide variety of industries.
8
For our purposes, a firm is treated as “producing” oil and gas discoveries the output from various
resources such as exploration and development expenditures the inputs. A simple Cobb-Douglas production model relates output Yi for firm i to inputs X
1
and X
2
over t
periods as follows:
9
Y
i
5 A
i
X
b
1
1
X
b
2
2
h 1
such that, by taking natural logarithms, one arrives at the regression equation lnY
i
5 a
i
1 b
1
lnX
1
1 b
2
lnX
2
1 e 2
where a
i
5 lnA
i
3 and
e 5 lnh
4 Assuming the requirements for regression are met for example, that e is a normally
distributed error term, the regression coefficients may be estimated. In a Cobb- Douglas production model, if the sum of the b slope coefficients equals 1, the
production model has constant returns to scale; if less than 1, decreasing returns to scale; and if greater than 1, increasing returns to scale Chiang, 1984, p. 414. If there
are constant returns to scale, size by itself does not affect efficiency.
For given quantities of X
1
and X
2
and given b slope coefficients, the estimated magnitude of the multiplier A
i
will proportionately affect the estimated level of Y
i
. The multiplier A
i
is an efficiency parameter Chiang, 1984, p. 416. The firm-specific intercept in the regression equation, a
i
, is the natural logarithm of the firm-specific multiplier A
i
. The estimate of the firm-specific multiplier A
i
is the natural anti-log of the estimate of the firm-specific intercept a
i
. If all firms are assumed in the estimation to share common beta slope coefficients, then for any fixed quantities of the inputs,
the ratio of the expected output for any two firms will be in the ratio of the estimated firm-specific multipliers for the firms. We follow the practice of comparing all firms
to the most efficient firm Cornwell et al., 1990. If Firm 1 has the largest estimated firm-specific multiplier of all firms, dividing each firm’s estimated multiplier by the
estimated multiplier of Firm 1 will produce an efficiency index expressing each firm’s productivity in terms of Firm 1. Firm 1 will have an index of 1.00. If another firm,
say Firm 2, has an index of, say, 0.625, then Firm 2 is expected to have output of only 62.5 of that of Firm 1 from the same inputs. Firm 2 may be said to be 62.5 as
Boynton et al. Journal of Energy Finance and Development 4 1999 1–27 7
efficient as Firm 1. We refer to our Cobb-Douglas regression derived index as “EFF.” By construction, higher rank on EFF is a sign of exploration efficiency relative to
lower rank on EFF. We employ the following Cobb-Douglas production model with firm-specific inter-
cepts, year specific intercepts, and year specific slopes
10
to model the relation between oil and gas discovered the output or dependent variable and resources employed
the inputs or independent variables: LNADD 5 INTERCEP 1 d
i
FIRM
i
1 g 89DUM89 1 g90DUM90
1 g 91DUM91 1 b
1
LNEXP 1 b
11
LNEXP 89 1 b
12
LNEXP 90
1 b
13
LNEXP 91 1 b
2
LNDEV 1 b
21
LNDEV 89 1 b
22
LNDEV 90
1 b
23
LNDEV 91 1 b
3
LNOUT 1 b
31
LNOUT 89 1 b
32
LNOUT 90
1 b
33
LNOUT 91 1 b
4
LAGEXP 1 b
41
LNLEXP 89
1 b
42
LNLEXP 90 1 b
43
LNLEXP 91 1 b
5
LAGDEV 1 b
51
LNLDEV 89 1 b
52
LNLDEV 90 1 b
53
LNLDEV 91 1 e
5 where
LNADD 5
natural logarithm of SFAS 69 additions and extensions to proved reserves in thousands of BOE converting 6 mcf of natural gas to 1
BOE. INTERCEP 5
the intercept for the regression equation and the estimate of the firm- specific intercept a
1
for Firm 1. FIRM
i
5 dummy variable equal to 1 for Firm i observation and equal to zero
otherwise for i 5 2–119; the sum of INTERCEP and the coefficient estimate for FIRM
i
, d
i
, is the estimate of the firm-specific intercept a
i
for Firm i. DUM
yy 5
dummy variable equal to 1 for year 19yy observation and equal to zero otherwise yy 5 89, 90, or 91.
LNEXP 5
natural logarithm of SFAS 69 exploration expenditures in thousands of dollars.
LNEXP yy
5 natural logarithm of SFAS 69 exploration expenditures in thousands
of dollars for year 19yy observation and equal to zero otherwise yy 5 89, 90, or 91.
LNDEV 5
natural logarithm of SFAS 69 development expenditures in thousands of dollars.
LNDEV yy 5 natural logarithm of SFAS 69 development expenditures in thousands
of dollars for year 19yy observation and equal to zero otherwise yy 5 89, 90, or 91.
LNOUT 5
natural logarithm of SFAS 69 oil and gas production in thousands of BOE.
LNOUT yy 5 natural logarithm of SFAS 69 oil and gas production in thousands of
8 Boynton et al. Journal of Energy Finance and Development 4 1999 1–27
Table 1 Cobb-Douglas Model Regression Estimates, Related Test of Coefficient Restrictions, and Calculated
Exploration Efficiencies for 119 Oil and Gas Firms for 1989–1992 Estimated
2-tailed Variable
coefficient t statistic
significance level Intercept
2 1.259
2 0.909
0.364 LNEXP
0.260 1.790
0.074 LNDEV
0.395 2.796
0.006 LNOUT
0.293 0.987
0.324 LAGEXP
0.007 0.040
0.968 LAGDEV
0.087 0.412
0.681 DUM89
0.890 1.554
0.121 DUM90
0.987 1.675
0.095 DUM91
0.444 0.815
0.416 LNEXP89
2 0.169
2 0.811
0.418 LNEXP90
0.191 0.962
0.337 LNEXP91
0.378 1.789
0.075 LNDEV89
0.284 1.117
0.265 LNDEV90
2 0.150
2 0.572
0.567 LNDEV91
2 0.145
2 0.575
0.566 LNOUT89
2 0.292
2 1.405
0.161 LNOUT90
2 0.425
2 1.987
0.048 LNOUT91
2 0.412
2 1.996
0.047 LNLEXP89
0.180 0.814
0.416 LNLEXP90
2 0.038
2 0.199
0.842 LNLEXP91
2 0.022
2 0.109
0.913 LNLDEV89
2 0.100
2 0.412
0.681 LNLDEV90
0.331 1.266
0.207 LNLDEV91
0.140 0.531
0.596 R-square value is .91; adjusted r-square value is .87.
BOE for year 19yy observation and equal to zero otherwise yy 5 89, 90, or 91.
LAGEXP 5
natural logarithm of SFAS 69 exploration expenditures for prior year in thousands of dollars.
LNLEXP yy 5 natural logarithm of SFAS 69 exploration expenditures for prior year
in thousands of dollars for year 19yy observation and equal to zero otherwise yy 5 89, 90, or 91.
LAGDEV 5
natural logarithm of SFAS 69 development expenditures for prior year in thousands of dollars.
LNLDEV yy 5 natural logarithm of SFAS 69 development expenditures for prior
year in thousands of dollars for year 19yy observation and equal to zero otherwise yy 5 89, 90, or 91.
Table 1 presents the estimated coefficients for the Cobb-Douglas regression for our sample of 119 oil and gas firms for 1989–1992. The model uses data from both
the current year and the prior year in each observation. Using 5 years of data 1988– 1992, we have 4 years of observations 1989–1992 for each firm or 476 total observa-
Boynton et al. Journal of Energy Finance and Development 4 1999 1–27 9
Table 2 Test of Coefficient Restrictions
F statistic p value
Test SCALE8992: Joint test that the coefficients on LNEXP, LNDEV, LNOUT, LAGEXP and LAGDEV sum to 1; that the coefficients on LNEXP91,
LNDEV91, LNOUT91, LAGEXP91 and LAGDEV91 sum to 0; that the coefficients on LNEXP90, LNDEV90, LNOUT90, LAGEXP90 and
LAGDEV90 sum to 0; and that the coefficients on LNEXP89, LNDEV89, LNOUT89, LAGEXP89 and LAGDEV89 sum to 0.
1.09 .36
Test EXP: Joint test that the coefficients on LNEXP89, LNEXP90, LNEXP91 are all 0.
3.03 .03
Test DEV: Joint test that the coefficients on LNDEV89, LNDEV90, LNDEV91 are all 0.
1.15 .33
Test OUT: Joint test that the coefficients on LNOUT89, LNOUT90, LNOUT91 are all 0.
1.72 .16
tions for the 119-firm sample for the regression. The R
2
value of .91 adjusted R
2
value of .87 suggests that the model captures much of the complexity of the exploration process. The model explains approximately 91 of the variation in LNADD, the
natural logarithm of the additions and extension to proved reserves. Table 2 presents tests of various restrictions on the estimation of the Cobb-Douglas
regression. One cannot reject at normal levels of statistical significance the restriction that the sum of the beta slope coefficients for each year equals 1, suggesting that
there are constant returns to scale Kmenta, 1986, pp. 412–422. Test SCALE8992 has a p 5 .36 . .10. One can reject the restriction that the coefficients on LNEXP
in each year, 1989–1992, are equal, suggesting that in one year a significant difference occurs in the coefficient on LNEXP and is allowed in the model by the use of year-
specific bs slopes. Test EXP has a p 5 .03 , .05. However, one cannot reject at normal levels of statistical significance the restriction that the estimated coefficients
on LNDEV in each year are equal. Test DEV has a p 5 .33 . .10. The same is true for the estimated coefficients on LNOUT. Test OUT has a p 5 .16 . .10.
Table 3 presents the Cobb-Douglas regression derived index of exploration effi- ciency EFF for each of the 119 firms in our sample ranked on EFF. The measure
EFF for each firm is the estimate of the firm-specific multiplier for each firm divided by the largest such estimated multiplier.
11
The measure EFF for a firm is a single value for 1989–1992 and represents the average efficiency of a firm during the 4-year
estimation period compared to the average efficiency of the firm that was most efficient. By construction, higher rank on EFF is a sign of exploration efficiency relative to
lower rank on EFF. In section 3, we compare rank on EFF with rank on each of four finding costs ratios and rank on each of two operating profit ratios.
3. Descriptive statistics for sample of 119 publicly owned oil and gas firms for 1990–1992
