You are on page 1of 14

Energy Vol. 5, pp. 617dM 0?

4&5442180/0701_0617/SCEWO
@ Pergamon Press Ltd.. MO. Printed in Great Britain

SHORT TERM CURTAILMENT OF NATURAL GAS:


ALTERNATIVE CURTAILMENT FORMS AND
EMPLOYMENT IMPACTS IN THE 1976-77 GAS EMERGENCYt

B. VON RABENAU

Department of City and Regional Planning, The Ohio State University, 289Brown Hall, I90West 17thAvenue,
Columbus, OH 43210,U.S.A.
and

WILPEN L. GORR
School of Public Administration, The Ohio State University, 202A Hagerty Hall, 1775 College Road,
Columbus, OH 43210,U.S.A.

(Received 2 October 1979)

Abstract-The 1976-77heating season was the coldest on record, forcing utilities throughout the Northeast
and Middle West to curtail their customers severely. Based on Ohio survey data, the employment impact of
these curtailments is estimated by industry and week. Several types of curtailment were used, including
appeals for voluntary conservation. As these curtailments allow customers a variety of responses, the study
differentiates among curtailment types and assesses the employment impact of each.
The impact estimates have been used in the design of an emergency curtailment system in Ohio. They
also provide useful information in non-emergency situations for utilities faced with a decision to increase
capacity and to expand customer hook-ups. As capacity is usually insufficient to preclude all shortages, it is
important that the cost of capacity increases be weighed against the cost of the sbortages they could avoid.

1. INTRODUCTION
The winter of 197677 was, in many parts of the U.S., the coldest winter on record. Below
average temperatures for the first four months of the heating season contributed significantly to an
already existing natural gas shortage. Ohio was one of the states worst hit: during the peak of the
crisis, public utilities were concerned about line pressure losses that would have forced the partial
closing of gas lines and the evacuation of residents; the governor declared a state of emergency in
January; public schools closed for several weeks in order to save natural gas; and the public utilities
curtailed most of their large customers to emergency levels.
During this emergency, the Public Utilities Commission of Ohio (PUCO) and the Ohio
Department of Energy (ODOE, formerly Ohio Energy and Resources Development Ad-
ministration) had no capability to model and forecast independently the consumption of gas
customers and particularly that of the industrial sectors. When customers had to be severely
curtailed, these agencies did not know how much gas savings would result or where layoffs
would occur, and thus they had little information on which to base their actions.
Following this heating season, a decision was made by-ODOE to obtain such an independent
forecasting capability. A model was developed which tracks and projects the gas consumption
and worker layoffs under alternative curtailment and temperature conditions. This model is
described in detail elsewhere.’ Here we focus solely on the employment impacts of the gas
shortage and the curtailments enacted during the 197677 heating season.
The analysis of the employment impacts is based on a survey of the large gas customers of
Columbia Gas of Ohio. Columbia Gas is one of only four major gas utilities in Ohio. It serves
roughly 40% of the state area, including all or part of 57 of Ohio’s 88 counties; and it is Ohio’s
second largest utility in terms of total gas sales. The survey of its customers was undertaken by
The Ohio State University for the PUCO immediately following the end of the 197677 heating
season.’ It contains, among other data, information regarding a firm’s layoff of workers, loss of
work hours, level of employment, seasonal gas consumption, base gas allocation, and alternate
fuel capabilities.

IThis paper extends results of a study for the Ohio Department of Energy undextaken by 0. Fisch, S. Gordon, W. Gorr,
R. Jacobs, J. Ludwig, K. Pearlman, and B. von Rabenau, all of The Ohio State University. Views and interpretations in this
paper are not necessarily those of the Ohio Department of Energy. Special thanks go to Harold Mba and Randy Crane
for their computational assistance, to Holly Mion for helpful suggestions and to Eric Tanner for tabulations of Federal fuel
statistics.

617
618 B. VON RABENAU
and WILPENL. GORR

Survey data of this type are of interest because extreme gas shortages are rare and because,
as a result, it is difficult to plan for them. They enable us to consider the following issues:
(1) Because there exist few data, it is customary to assume that workers are laid off in the
same proportion in which gas falls short of normal requirements (adjusting for conversion
capability and weather conditions). The survey data allow us to test this proportionality
assumption implied in particular in input-output-based impact assessments.
(2) While extreme and prolonged shortages, as were experienced in the survey winter, are
rare, less severe shortages are quite common. This statement is even true today when utilities,
faced with an increasing supply of natural gas, once again accept new customers. Shortages are
part of an efficient delivery system, which has a capacity somewhat below the peak demand
generated by extreme temperature conditions. To determine a desirable level of shortages one
must first assess the adverse impacts from gas shortages.
(3) Unexpected short run gas shortages cannot be managed through prices. Prices can only be
changed after lengthy public hearings and must be considered fixed in the short term. The main
management tool then becomes the curtailment or rationing of gas, or appeals for voluntary
conservation.
Very little is known so far about customer responses to differing curtailment designs. An
empirical study by Peck and Doering shows that an appeal for strictly voluntary conservation is
largely ineffective in reducing the consumption of gas among residential customers.3 They
conclude that such appeals must be combined with price increases to be effective. Their study,
however, does not deal with more coercive curtailment forms of non-residential customers, nor
does it pertain to short term emergency situations such as are considered in this paper.
Curtailments are usually assumed to fully determine the customer response. As a result, this
response is rarely studied. In reality, however, almost all curtailments allow for some response
flexibility either because compliance cannot be or is not monitored, because penalties are
inadequate, or because curtailment rules allow for flexibility in the timing of cutbacks. Hence,
curtailments do not determine gas consumption via a strict accounting identity. Rather, the impact
of curtailments on gas consumption and employment must be treated in a behavioral model, thus
making emergency planning more difficult than it would be otherwise. Such a model is proposed in
this paper and the response of the Columbia Gas customers to different types of curtailments is
analyzed.
(4) A number of programming models has been designed to evaluate energy allocation,
development, and storage strategies in the face of energy shortages.4,’ Such models assume
information on the cost of energy shortage that is not currently available. Survey data such as
those discussed here can provide a start for a data base to assess the cost of natural gas
shortages. Of course, only the direct employment impact on the curtailed firms was assessed.
There may have been additional layoffs caused indirectly by the gas curtailments. Also, some of
the work losses were later made up in overtime. Still, given the extent of the shortage, which
resulted in the curtailment of almost all gas for firms that employ 25% of the work force in their
area, it is surprising to find that the shortage contributed only half a percentage point to the
unemployment rate in January and February of the survey year.
In the following section, we describe the curtailment system in effect for Columbia Gas
customers. In Section 3, we provide a descriptive analysis of selected survey results through
cross-tabulations. Section 4 develops a behavioral model of a firm’s employment response to
curtailments. Worker layoffs are regressed on the different curtailments and weather conditions,
and estimates are reported by industry. A final section provides a summary and conclusions.

2. CURTAILMENT PROCEDURES
Prior to 1974, Columbia Gas did not curtail its customers except under extreme weather
conditions when high residential demand threatened a drop in line pressure and endangered the
safety of line operations. For such peak-day curtailments, Columbia Gas relied on its special
contract customers which receive a high volume and stable load of gas at decreased rates.
Contrary to the practice of many other utilities, Columbia Gas’ special contract customers held
(and still hold) a firm seasonal gas commitment and hence were not classified “interruptable”.
Because on the average only 2 or 3 days annually were peak-day curtailed, most of these firms did
not have a full alternate fuel capability.
Short term curtailment of natural gas 619

Since December 1975, Columbia Gas has curtailed its customers through both peak-day and
volumetric curtailments. The two curtailment types vary in their use and purpose, in the extent
to which compliance can be monitored, and in the decision freedom they leave the gas
customer. Both were used by Columbia Gas during the 1976-77 heating season; in addition, of
course, firms were asked by the Governor of Ohio for voluntary cutbacks. We consider each of
these in turn.
(1) Volumetric curtailments limit the total volume of gas a customer may use over the winter
season. They are imposed on commercial and industrial users so that the utility company may
stay within the seasonal gas limit for which it has contracted with the transmission company.
However, rather than expressing the volumetric curtailment to the curtailable customer as a
seasonal gas constraint, it is expressed in the form of monthly, non-binding levels of allowable
gas consumption. These levels are set mainly in response to the gas consumption by non-
curtailable higher priority gas users, in particular, residential customers. While the monthly
levels of allowable gas are not binding, the summation over the season of these monthly levels
represents a seasonally binding limit. Obviously, a customer who complies with each month’s
constraint also complies with the seasonal constraint. But customers have the freedom to
follow other consumption paths as long as they stay within the seasonal limit. One difhculty for
the customer in following such other paths is, however, that he does not know the future
monthly levels of allowable gas or, for that matter, the seasonal total constraint.
For the purpose of volumetric curtailments, Columbia Gas has given each firm a base period
allocation of gas, representing the ration in the absence of volumetric curtailments. This
allocation is based on the firm’s gas consumption during a historical, uncurtailed base period.
The seasonal base allocation is further broken down by month and by end-use into non-
substitutable, substitutable and boiler gas allocations. Volumetric curtailments are expressed as
a percent of this base allocation. Curtailment levels vary among customers according to priority
class and gas end-use. In particular, customers are divided into large, intermediate and small
customers, and large and intermediate customers are further divided into industrial and
commercial gas users. A customer is large if its monthly base allocation exceeds 1000MMCF.
The customer is classified “commercial” if more than 50% of its gas is used for space heating
purposes; it is called “industrial” otherwise. These terms in no way imply a particular industry
affiliation.
Using a complex priority system, large customers are curtailed at higher levels than medium
or small customers. Despite the extreme emergency of the 197677 heating season, the
curtailment of intermediate customers was relatively small, restricted to 30% during February
and to 10% during January of 1977. Because most small customers had not yet been assigned a
base allocation, they were never curtailed. Most of the curtailment burden fell, therefore, on the
large customers. Their base allocation represents 82% of the combined base gas of small,
intermediate and large users. Thus, even if the small and intermediate users had been curtailed
at a higher rate, the potential gas savings would have been small. Because almost all
curtailments fell on the large customers, it seems justified to restrict this study to these
customers.
The volumetric curtailments imposed on large customers are shown in Table 1. This table
shows the curtailment for each week of the heating season of large commercial and industrial
users. The curtailments for industrial users are imposed by end-use and must be summed, after
weighting each of them by the proportion which the end-use represents in the total base gas of
the user. Column 5 shows such weighted curtailments using industrial users in SIC class 3&39
as an example.
(2) Peak-day curtailments do not affect the seasonal gas consumption constraint, but they
limit the amount of gas a hrm may use on a particular day, typically to emergency levels of as
little as 10% of the base allocation. As mentioned earlier, such curtailments are imposed to
avoid losses in line pressure that result when the rate of gas consumption exceeds the rate of
possible re-supply. Compliance with such curtailments is difficult to monitor, and would require
frequent readings or special meters recording the gas consumption on a continuous or daily
basis.
Peak-day curtailments were imposed only on large customers. There are two peak-day
curtailment periods shown in Column 6 of Table 1. During the first period (1619 January 1977),
EGY Vol. 5. No. 7-D
620 B. VON RABENAU and WILPENL. GORR

Table 1. Curtailment as a percentage of the base allocation and the week days lost per loo0employees
(1976-77).

T- Volumetric curtailments (umber of


fork days Work days lost per
(umber
of
I- ndust al with 1000 employees
I SIC 30-39
week leak-day
Non- rota1 Com- lrtailmen
.ubsti- ubsti ioil- SIC ercia
.utable utabl er IO- 39 Industria Zommercia

(1) (2) 131 (4) (5) (6) (7) (8)


10 100 40.7 15.0 4.2
10 100 40.7 18.7
10 100 40.7 24.0 65.6”
10 100
100
40.7
40.7
2
40
27.0
30.7
5:s
6.6
10

10 65 100 40.7 40 30.7 6.6


10 65 100 40.7 40 33.8 6.6
10 65 100 40.7 40 29.9 5.5
14 69 100 44.5 42 31.7 5.5
50 100 100 74.0 50 43.0 6.6

11 50 100 100 74.0 50 56.0 6.6


50 100 100 74.0 50 242.5 5.5
:: 100 100 74.0 50 254.3 90.2
14 :08 100 100 88.6 582.4 260.3
15 85 100 100 91.8 ;z 328.8 324.5

16 100 100 91.8 85 161.8 61.7


17 :: 100 100 91.8 85 92.5 52.0
18 48 100 100 73.0 46 44.5 23.8
10 86 100 56.7 38.2 15.8
:: 10 70 100 42.8 z 29.3 14.9

10 100 42.8 25.6


:: 10 100 42.8 18.9

Total 2159.4 930.3

only the 483 large customers with special contracts were curtailed and, during the second period
(27 January-8 February 1977), the 1650 of the 2364 large customers were curtailed, for whom
the industry affiliation was then known. As an illustration of the week by week employment
impact of both peak-day and volumetric curtailments, we show in Columns 7 and 8 the work
days lost per 1000 workers in industry classes SIC 30-39.
(3) A final form of curtailment was the declaration of a state of emergency by the Governor
of Ohio on Sunday, 23 January 1977. This declaration, which was not lifted through the
remainder of the winter, was coupled with appeals for voluntary conservation and cutbacks in
working hours, in particular of stores and commercial establishments.

3. SURVEYRESULTS
Of the 2364 large customers of Columbia Gas, a total of 637 or 27% responded to the mail
survey. Of these responses, 130 were from public schools. They are deleted from the analysis
because their responses were often difficult to interpret.’ The remaining 507 responses represent
28% of the approx. 1835 large non-school customers. Also, they hold 37% of the base gas of all
large customers and 3% of the base gas of non-school customers.
Table 2 summarizes, by industry, the survey findings relating to the employment impact of
the 1976-77 winter gas shortage. We note that for 384 firms complete information did exist with
regard to their base allocation gas, November 1976employment, industry affiliation, and layoffs.
These firms, at the beginning of the 1976-77 heating season, employed 133,512 workers.
Assuming that the sample exhibits the same bias towards large employers as it does towards
large base allocations (i.e. the 507 responses represent 3% of the workers of all large
non-school customers), the total work force of large non-school customers is estimated to be
between 350,000and 420,000 workers. This represents 1%23% of the roughly 1.8 million workers
in the Columbia Gas service area. If the schools are included, then more than every fourth worker
was employed by a large gas customer, and hence every fourth worker was directly affected by the
curtailment of gas to these customers.
Of the 384 firms in Table 2,119 firms had a total 207 layoffs related to the gas shortage. Most
of these layoffs occurred to manufacturing firms with Standard Industrial Classifications (SIC)
Table 2. Selected employmentimpact measures,by industry,for large customersof ColumbiaGas.

SIC Number Number of Number of Total Average Average Total ?ercentage Estimated
code of firms Industry firms Number of length ercentage umber of of rate of
employees with layoffs of of ,ork days >otential unemploy -
layoffs layoffs workers lost rork days ment
(number laid off lost
of days)

(1) (21 (4) (5) (6) (7) (8) (9) (101

26 580 7 58 57.2 904.8 0.95 0.83


038 07 0.0 0.0
645 11581 2i.3 53.2 12792.3 0.94 0.64
198 14459 :z 1:: 20.4 46.8 196957.8 2.26 1.48
7149 2 8 88.2 2655.0 0.32 0.04
a: 49 30 : 8 13.4 48.3 3666.5 0.64 0.31
5257 0 0.0
i 1107 3 3 49.7 21.6 996.8 0.77 ::066
11 26596 1 1 6.0 11.5 38.5 0.0
5 1007 0 0.0 8::

I 47.8 218091.7 1.40 0.95


Total 384 153512 1 119 207 21.8

This table excludes firms with incomplete information on base allocation gas, November employment,
industry affiliation, or layoffs.
622 B. VON RABENAU
and WILPENL. GORR

30-39. In that industry, 43% of all firms had layoffs. From Columns 6 and 7 it is seen that the
typical layoff involved almost 50% of all workers of a plant and lasted 21.8 days on the average.
Over the entire winter season, the gas shortage resulted in the loss of 218,092 workdays among
the surveyed firms. Since these firms had 133,512 workers who could potentially have worked
15.7 million workdays during the winter season, approx. 1.4% of the potential workdays were
lost to workers in these firms. As Column 9 of Table 2 indicates, this number is substantially
higher among manufacturing firms in SIC class 30-39. Also, it must be remembered that the
major impact from the gas shortage occurred during January and February 1977. For example,
during the week from 30 January-5 February, 6.6% of the potential workdays were lost for the
surveyed firms, and this number increases to 10.6% if only industrial firms in SIC class 3&39
are considered.
It is of interest to estimate the extent to which the gas shortage contributed to the official
unemployment in the Columbia Gas service region. In official unemployment statistics, a person
is counted as unemployed only if the person for the entire week (a) is on layoff or (b) does not
work but is looking for work. (For a complete definition, see Ref. 6.) Thus, unless a person is
laid off an entire calendar week none of the workdays lost in that week should be counted
toward official unemployment. As a rough approximation, if the average layoff has a duration of
more than seven days, say x calendar days, then the proportion of workdays that count towards
official unemployment is (x -7)/x. Using this approximation yields the estimated rate of
unemployment caused by the gas shortage among workers of large gas customers, shown in
Column IO of Table 2.
Column10 shows that the gas shortage added, during the five winter months, an average of
almost one percentage point to the unemployment rate among workers of large gas customers.
Actually, 81% of the workday losses occurred during January and February of 1977. During
each of these two months, therefore, the gas shortage added 2 percentage points to the
unemployment rate of these workers. Finally, to the extent that these workers represent only
about 25% of all workers in the service area, and to the extent that there was no additional
indirect impact from the shortage on other firms, we conclude that approximately half a
percentage point of the official unemployment rate of 6.5% in January and 7.5% in February of
1977 was due to the gas shortage.
For later reference, Table 3 shows the capability of firms to convert gas demand to alternate
fuels as a percentage of the base allocation of gas. These estimates are based on a survey
question that determined the capability installed since 1970. A subsequent telephone check with
respondents showed that little capability existed prior to 1970. Where it did exist, this had
typically been indicated on the survey form and is thus reflected in the estimates of Table 3.
The estimates of the installed conversion capability were also checked against the results of a
Federal survey that determines annually the alternate fuels actually used.’ Unfortunately, this

Table 3. Industry conversion capability (total industry gas convertible to alternate fuels as a percentage of total
industry winter base allocation), by industry and priority class, for large customers (of the Columbia Gas
Company).

Gas priority class

l-Digit Large industrial Large commercial Total


SIC
code conversion Number of conversion Number of conversion Number of
capability firms capability firms capability firms

0 65.3% 9 24.8% 12 37.6% 21


: 30.1
38.0 602 3E 6
1 35.8
30.1 663

3 44.2 170 15:2 24 43.7 194


4 46.8 3 19.1 12 36.0 15
5 21.6 16 19.3 20 21.1 36
6 8.4 8 8.4 a
7 69.9 3 14.2 7 43.8 10
8 49.4 3 22.7 7 46.7 10
9 8.3 1 10.1 4 8.0 5

Total 37.3% 267 18.2% 101 36.6% 368

This table excludes firms with incomplete information on the base allocation of
gas, industry affiliation or conversion capability.
Short term curtailment of natural gas 623

survey does not allow a breakdown by SIC and priority class. Also, the alternate capability used is
usually below that installed. However, if one makes the reasonable assumption that under the
severe conditions of winter of 197677 all installed conversion capability was used one finds an
overall conversion capability of between 35 and 40% which is similar to the estimate reported in
Table 3.
From Table 3 it is seen that the conversion capability varies by industry and priority class.
Large industrial customers generally have a higher conversion capability than commercial
customers with a predominant demand for space heating. Regression analysis shows that much
of this variation can be explained by previous curtailments.’ Industries or firms that
experienced higher curtailments in the past (and that can expect relatively higher curtailments
in the future) have acquired a greater conversion capability.

4. REGRESSION RESULTS
In this section, we estimate weekly workday losses due to the energy shortage as a function
of volumetric and peak-day curtailment levels and degree days. The analysis is undertaken at
the level of industrial sectors using as observations the 21 weeks of the winter season.
Alternative levels of aggregation are possible: for example, the analysis could be undertaken
using observations on days or on individual firms. However, the chosen higher level of
aggregation avoids autocorrelation problems that result from the lumpiness of layoffs, observed
in Table 2.

Behavioral aspects of curtailments and layofs


Peak-day and volumetric curtailments and the Governor’s appeal for conservation differ in
the degree of freedom they leave the individual gas customer with regard to his weekly gas
consumption, The impact of each type of curtailment on weekly layoffs or loss of workdays
would, therefore, be expected to differ also.
Peak-day curtailments require an immediate and daily compliance that seems fairly rigidly
enforced (though for the majority of firms gas is metered only once a month). We would
therefore expect a relatively strong layoff response, exceeding that resulting from comparable
volumetric curtailments.
Volumetric curtailments do, as was discussed earlier, require compliance with their unknown
seasonal total but not their current levels. Current curtailment levels therefore have the
character of guidelines or orientation marks. How much gas a firm uses and how many workers
a firm lays off at any moment in time depends on its expectations regarding the seasonal
curtailment total. To the extent that current information influences the expected seasonal total
curtailment, layoffs may be a function of current volumetric curtailment levels. But this need
not be the case. In particular, assume firms have perfect foresight with regard to their seasonal
total volumetric curtailment. In this case one would expect current curtailment levels to have
no effect on current employment levels. The various curtailments in effect during the winter
season are then simply a cumbersome way of presenting an overall constraint on the total
winter gas consumption. In the special case where none of the other decision parameters of a
firm depends on time, the firm will simply produce a constant level of output during the heating
season (possibly modified by a discount factor), lay off a constant number of workers each
week, and use a constant amount of gas each week so as to arrive at the seasonally allowable
gas by the end of the season. Weekly layoffs would not be explained by weekly curtailments.
In contrast to the above case, assume a firm has imperfect foresight and believes that the
current curtailment level will last to the end of the season. As in the previous case, this firm
maximizes profits designing a gas consumption and employment plan subject to the perceived
seasonal curtailment constraint. Of course this firm must revise its plan each time the
curtailment level is changed during the winter. Considering again the special case where alJ
decision parameters are time invariant, it is easily seen that such a firm, during each week, will
use the amount of gas allowable under the curtailment in effect that week. Put differently, this
firm will behave as if it had to comply in each week with the curtailments in effect that week.
One would expect a strong relationship between weekly layoffs and concurrent curtailment
levels, a conclusion opposite to that obtained by the perfect foresight model.
Appeals for voluntary curtailments by the Governor are likely to have little impact on gas
624 9. VONRAEIENAU and WILPENL. GORR

consumption and layoffs. In particular, those firms already subject to mandatory curtailments
should be expected to show no additional impact. At best, one would expect some impact in the
commercial sector where relatively new firms were subject to mandatory curtailments, where
curtailment levels were low, and where public pressure and exposure may result in con-
servation measures.
A firm’s capability to convert to alternate fuel complicates the relationship between layoffs
and curtailments, in particular, volumetric curtailments. In the short run, which is considered in
this paper, this capability is fixed, and the cost of switching to alternate fuels is small. A !irm
that can switch (or has permanently converted) a certain percentage of its natural gas demand
to alternate fuels will remain unaffected by gas curtailments up to that percentage. In this case
the true relationship between layoffs and curtailments is kinked as shown in Fig. l(a).
Curtailments up to the conversion capability will not result in any layoffs but larger curtail-
ments will. Estimating a single regression line for both regions will underestimate the “true”
slope above the conversion capability but will overestimate it below that capability. This is of
course only a problem if observations exist in both regions. In the winter of W&77,
curtailments throughout the winter were above or only slightly below the industry’s average
conversion capability so that the estimated slope would be equal to theme slope in this region
if all firms had the industry average conversion capability. Of course this estimated equation
would not apply to the lower curtailment region.
When many firms with different conversion capabilities are aggregated to an industry, then
the relationship between curtailments and layoffs will have many kinks that can be treated as a
smooth curve provided the number of firms is sufficiently large, see Fig. l(b). This relationship
can be estimated by splicing together linear functions with increasing slope (which in this case
would result in degree of freedom problems), or by raising the curtailment measure to a power
larger than one.
Cold weather could be a final contributing source to layoffs. However, while temperature

Layoff

conversion base curtailment


capability

(a)

Layoff

relationship

base curtailment
of observed
curtailments

(b)
Fig. 1. Relationship between layoffs and curtailments.
Short term curtailment of natural gas 625

plays a large role in explaining gas consumption, in particular for space heating, it is not clear
that its role should be equally prominent in explaining layoffs. When outside temperatures fall
extremely low, capacity constraints in heating equipment may require layoffs if inside tem-
peratures cannot be maintained at comfortable levels. However, a layoff does not realize
increasing gas savings as outside temperatures drop. This is so, because inside temperatures
must be maintained at a safe plant maintenance level, and hence savings from a shutdown do
not increase once the outside temperature falls below this level. Therefore, any association
between layoffs and the outside temperature is expected to be weak.

Variables
The model regresses the per cent WL of a week’s potential workdays lost to layoffs on (a)
weekly degree days (DDW), (b) various measures of the volumetric curtailment level (CTPW,
CTPW2) in effect that week, (c) alternative measures of the week’s peak-day curtailment
(PEAKl, PEAK2), and (d) a variable designed to capture the impact of the Governor’s
emergency declaration (EMER). Each of these variables is briefly discussed.
WL is the dependent variable, the workdays lost during a week due to the gas shortage, as a
per cent of that week’s potential workdays. The potential workdays are computed assuming five
workdays for each of the sector’s November 1976 employees (fewer workdays if the week
contains a holiday).
CTPW represents a sector’s weighted average volumetric curtailment for a given week,
computed as a per cent of that sector’s base allocation. The weighting takes into account
changes in the curtailment level during a week. Also, industrial users actually face three
different curtailment levels, one each for their non-substitutable, substitutable and boiler gas
allocation. For these sectors therefore, CTPW represents the weighted average of these
curtailment levels, the weights being the amount of gas allocated to the sector in each category.
CTPW2 equals CTPW, squared. It is introduced to allow for the expected convexity in the
relationship between WL and CTPW discussed in the previous section.
DDW, the weekly degree days, is a measure of how cold a week is. It is the summation over
the week of the difference between 65” and the outside temperature.
PEAK1 and PEAK2 are two alternative measures of a sector’s weekly exposure to peakday
curtailments. Each measure is the product of two terms: the percentage of a week’s workdays
for which the curtailment is in effect, and the proportion of the sector’s base gas held by the
firms subject to the curtailment. Hence, PEAK1 and PEAK2 assume values between 0 and 100.
Since it takes some time to notify firms of a peak-day curtailment and since in turn it takes time
for firms to notify their workers of a layoff, several alternative lags were assumed in computing
the percent of a curtailment that falls into each week. In particular, PEAK1 assumes a one day lag
and PEAK2 assumes a two day lag.
EMER is a measure of the state of emergency. It could be designed as a dummy variable that
is zero prior to week 13 and unity thereafter (as the state of emergency was not lifted after its
imposition on Sunday, 20 January 1977). However, when tried, this measure was never
significant. It was felt that this was possibly the result of a decline in the impact on customers
of the initial dramatic appeals by the Governor for voluntary gas savings and cutbacks, appeals
that were made in the first few days of the emergency. Therefore, an alternative measure was
tried that reflects this decline. In particular, EMER is given (for ease of comparison with the
other variables) a value of 100 for the first week of the emergency, and then its value is allowed
to decrease geometrically for the following weeks with constant ratio of 1 : 2.

Regression estimates
Separate equations for the workday losses WL are estimated for the agricultural sector (SIC
00-09), manufacturing firms (SIC 2&29 and SIC 30-39), retail and wholesale firms (SIC 50-59)
and the combined non-manufacturing sector (SIC 00-19, 40+). For each of these industries
estimates are reported separately for the industrial and commercial customers, and for the two
combined.
The estimated equations are shown in Table 4. Because there are only 21 observations for
each sector, degree of freedom problems prevented estimating a single equation containing all
variables. Two equations are reported for each sector. These equations do not vary among
626 B. VON
RABENAU
and WILPEN
L. GORR

Table4. Regressioncoefficientsfor equationspredictingWL (per cent of potentialweekly

Industry Priority Class Equation Constant CTPW CTPWZ


(SIC Code)

A riculture industrial ,o. oa6*


(10-09)

commercial -1295.47* 52.53” 0.408’


2.30 1.67’
total - 678.93” 20.44Sa 0.215’
31.55 1.2Q0

industrial 15.06 0.013’


[:I - 15.42 1.207*
commercial (3) 30.90 0.004
(4) la.74 0.117
total (3) 14.19 0.0131
(4) - 15.99 1.217*

Manufacture industrial (3) 13.05 0.025”


(30-59) (4) - 59.72 2.679’
I
commercial - 20.62 0.022’
!: j - a3.37? 2.267*
I
total (3) 13.16 0.025”
(4) - 56.32” 2.635*

Retail, industrial - 12.48 0.008


wholesale - 15.96 0.824
(50-59) -t
commercial 14.28 0.002
11.24 -0.227
total 0.005
: 1x: / -0.471

Total industrial (3) 15.84 0.024*


manufacture (4) - 54.00° 2.562’
(20-39)
commercial - 13.67 0.020*
((43j - 69.59+ 1.978*
total (3) 13.44 0.024*
(4) - 56.64+ 2.605*

Total non- industrial (3) 0.002


manufacture (4)
(00-19, 40*)
commercial (3) D.OOO
(4)
total (3) 3.002
(4)

*significant at .Ol level


+significant at .05 level
‘significant at .lO level
Short term curtailmentof naturalgas 621

workdays lost), by industry and priority class. (Au coefficients in per 100 of variable.)

DDU PRAKl PEAK2 EMER D-W F ti2

2.40 5.2 0.38


::f$*. 0.125 2.01 3.2 0.25

.OQl 0.67
.413+ 0.979 :-::. ‘Ct . 0.45

-.lOS 1.25 0.51


.105 0.607 0.87 ::‘z 0.25

1.612” 2.50 53.5 0.85


1.506* 0.899* 2.13 65.3 0.91

1.87 191.9 0.95


:21.
. :::* l 2.536” 1.95 232.4 0.97

2.215” 77.9 0.88


2.091. 0.908* :-::. 106.0 0.94

9.534. 2.04 196.6 0.95


8.571’ 2.015* 2.70 233.5 0.97

2.96 87.9 0.90


E:*
. l
1.491. 2.15 91.1 0.93

9.076* 2.03 221.2 0.96


8.355* 1.927’ 2.62 279.5 0.98

LO. 333’ 2.37 117.0 0.91


11.458* 2.066+ 2.98 100.5 0.94

0.384 2.31 0.06


0.584” 1.748* 1.71 7x 0.92

5.535. 3.00 0.94


5.691’ .0.243 3.10 ‘% . 0.93

8.317” 2.02 206.2 0.95


7.604* 1.836* 2.67 270.4 0.98

5.426’ 3.07 133.3 0.93


5.103” 1.632* 2.46 178.2 0.96

8.175’ 2.03 237.4 0.96


7.499* 1.752’ 2.58 336.7 0.98

2.87 123.4 0.92


:-:;,‘:
. 0.394’ 2.09 290.6 0.93

0.133* 1.59 4.6 0.27


-0.008 0.250’ 0.42 19.3 0.73

0.784* 2.17 86.1 0.94


0.901’ 0.131° 2.00 105.6 0.89
628 B. VON RABENAU
and WILPENL. GORR

sectors except for agriculture, and except for the PEAK variable. Individual sectors contain
either PEAK1 or PEAK2, depending on whether a one or a two day lag worked best for the
sector.
We first report the results for the agricultural sector, for which two alternative equations
were estimated, namely,

WL = a0 + a,CTPW + a+ZTPW’ + a3DDW, (1)

WL= b. + b,CTPW + b2DDW+ b,EMER. (2)

We find the following results:


(1) Very few firms in this sector were peak-day curtailed. When a peak-day measure was
included in the equations it was not significant, and hence it was dropped altogether.
(2) Degree days are highly significant but negatively related to layoffs. The colder the weather
the fewer people are laid off. At first this is a somewhat surprising result. However, it is easily
explained by the fact that most firms in this sector are greenhouse operators. Colder weather is
likely to require greater crop protection and maintenance, and hence leads to the recall rather
than the layoff of workers.
(3) The state of emergency declaration is, in contrast to all other sectors, never the source of
layoffs.
(4) Layoffs increase with volumetric curtailments, though unexpectedly, at a decreasing rate
(CTPW2 enters significantly negative). A 100% volumetric curtailment results in the layoff of
28% of the workers (in the combined industrial and commercial sectors) which is more than in
any other industry.
For the non-agricultural sector, two alternative equations were estimated, viz.

WL = ao + a,CTPW’ + azPEAKi, (3)

WL = b,, + b,CTPW + b2PEAKi + b3EMER, (4)

where i equals either 1 or 2, depending on the sector. Neither of these equations contains DDW
which in separate estimations was found not to be significant in all but one instance. In the first
equation CTPW enters in quadratic form and in the second it enters linearly. This will allow
inferences regarding the convexity in the relationship between WL and CTPW hypothesized
earlier. A brief discussion of results follows.
(1) Peak-day curtailments are, in contrast to the agricultural sector, significant in almost all
sectors both for industrial and commercial firms.
(2) Only for manufacturing firms are current employment levels explained by current
volumetric curtailment levels. These firms adjust their employment levels as indicated by the
imperfect foresight model, reacting to current information rather than only to a seasonal total
curtailment level. On the other hand, CTPW is not significant for non-manufacturing firms.
These fhms may nevertheless respond to volumetric curtailments though this response would
be more indirect. They may lay off a constant number of workers over the season in response
to an anticipated seasonal total curtailment; or they may shut down once the season’s allocated
gas is used up, to name but two possible response patterns.
(3) EMER is significant in all sectors. It contributed generally between one and two
percentage points to total layoffs and was, by implication, a considerable factor in gas cutbacks.
However, EMER is correlated with CTPW and PEAK3 and therefore has been dropped from
Eq. (3).
(4) Equations (3) and (4) strongly suggest that the relationship between curtailments and
layoffs, is indeed convex, and that the intercept is zero as shown in Fig. 1. As the level of
curtailments increases, the number of firms with still adequate conversion capability declines
and marginal layoffs increase. In this case one would expect a significant negative intercept if
CTPW enters linearly into the equation (as in Eq. (4)) or if the estimation is based only on
observations with large curtailment levels (as in the case for this winter). Also, one would
expect a zero intercept if CTPW is raised to some appropriate higher power (as in Eq. (3)) that
Short term curtailment of natural gas 629

reflects the convexity in the relationship between CTPW and WL. The estimates shown in
Table 4 confirm these expectations for almost all sectors. The intercept for Eq. (4) is negative
throughout and frequently significant. The intercept for Eq. (3) is not significantly different from
zero (except in one case in which CTPWZ is not significant). This also suggests that Eq. (3)
yields reasonable layoff estimates for smaller volumetric curtailments than those included in
our sample. (Obviously, the same should not be expected for Eq. (4).)
(5) It is of interest to compare for the manufacturing sectors the employment impacts of
peak-day and volumetric curtailments of similar magnitude. Since customers must reduce gas
consumption during peak-day curtailments to emergency levels of 90% of base allocation, we
compare the weekly workday losses from a full week of peak-day curtailments (PEAKI’ = 100)
to those resulting from a 90% volumetric curtailment (CTPW = 90). As might have been
expected, given the immediate compliance required for peak-day curtailments, their employ-
ment impact is twice to three times that of equivalent volumetric curtailments.
The generally high adjusted R* for almost all sectors suggests that the estimated equations
will do well in forecasting employment impacts from gas shortages. Impact forecasts are often
based on proportionality assumptions, such as are implied in the use of input output tables. How-
ever, our results suggest that while a considerable percentage of the work force was laid off this
percentage is much smaller than the per cent of the gas curtailed, even allowing for a sector’s
conversion capability. Moreover, monthly utility data suggest that this is not the result of
noncompliance with curtailments. The total gas consumption in each sector is almost always
just above or below its monthly curtailed base, even though this curtailed base is not a binding
constraint.
To be specific, consider the combined commercial and industrial manufacturing sectors (SIC
2,3; total). At the height of the emergency, virtually all gas was curtailed except about 10% of
plant maintenance gas. But the combined effect of all curtailments was to lay off only about
11% of the workers. The discrepancy between the 90% gas curtailment and the 11% layoff of
workers can be reduced somewhat by several adjustments.
First, we allow for the fact that generally only production workers are laid off (about 70% of
all manufacturing employees in Ohio). As a per cent of the production workers layoffs increase
to 16%. Second, we allow for the fact that the manufacturing sector in normal winters is
estimated to use only 80% of its uncurtailed base gas. As a per cent of this normal demand the
shortage then reduces to 87.5% (firms receive 10% of their base gas as plant maintenance gas
but demand 80% of base gas, leaving a shortage of (70: 80) x 100% of demand). Finally, the total
manufacturing sector has a conversion capability of approx. 37%, reducing the gas shortfall
from 87.5% to about 50% of demand.
After these adjustments it appears therefore that the short run gas elasticity of production
employment in manufacturing is 16 : 50 or about 1 : 3. A reduction in the effective gas supply
(that cannot be converted to alternate fuels) by 3% results in the layoff of about 1% of
production workers. Still, this is a much smaller than proportional impact. This should not be
surprising, however, and is largely the result of an aggregation of activities (possibly even at the
establishment level), some of which may not use any gas at all. Such activities may well
continue to operate when others, even in the same establishment, do not-leading to an
overestimate of the short term dependency of employment of gas. An extreme example would
be an establishment with two identical production lines housed respectively in buildings heated
by gas and electricity. Presumably, in this case a 100% curtailment results in the layoff of only
50% of the workers.

5. CONCLUSION
In this paper we have used data from the 1976-77 heating season to estimate the employment
impact of gas curtailments. Considering the severity of the curtailments during that winter and
the limited alternate fuel capability then in existence, the employment impact was relatively
small. We estimate that in manufacturing a 3% cutback of gas beyond existing conversion
capability results only in the layoff of 1% of the production workers. Obviously, this estimate of
layoffs would require upward revisions if the conversion capability or conservation efforts were
underestimated. However, we do not believe this to be the case and suggest that an assumption
630 B VON ~BENAU and WILPENL. GORR

of strict proportionality of layoffs and gas curtailments would overestimate the impact of gas
shortages considerably.
Regression estimates allow us to attribute layoffs to their individual sources. We find that
peak-day curtailments have an impact on layoffs twice to three times that of equivalent
volumetric curtailments. This is not surprising as peak-day curtailments require immediate
compliance and thus do not allow the firm a curtailment response tailored to its individual
needs. Hence, if curtailments are necessary, their cost to society can be reduced through the
careful scheduling of volumetric curtailments that avoids sudden emergency rationing similar to
peak-day curtailments.
The regression equations presented here have several uses. Earlier estimates have been
incorporated into a computer model at ODOE,9 designed to track the gas savings and
unemployment impacts from gas curtailments by the utilities or emergency curtailments by the
governor. The model forecasts these impacts for alternative curtailment plans and weather
scenarios,” and unemployment impacts serve as a criterion in deciding among alternative
curtailment plans. The equations may also be used in long term planning issues facing the
utilities. Currently, utilities explore the allocation of their increasing supply of gas. Even with
such an increasing supply it would not be efficient to avoid all short term shortages. However,
the cost of such shortages in form of layoffs and loss of output must be balanced against the
benefits from service expansion.

REFERENCES
0. Fisch, S. Gordon, W. Gorr, R. Jacobs, T. Ludwig, and B. v. Rabenau, Development of an Energy Allocafion Model to
Minimize Adverse Socio-Economic Impact During Periods ofSevere Energy
_. Shortage, 3 Vol. The Ohio State Universitv.I
Columbus, Ohio (1978).
M. A. Walters, K. Kelly,andT. Bydolek, The Emergency Purchase, Transfer, and Se/j-He/p Programs. The Ohio State
University, Columbus, Ohio, (1977).
A. E. Peck and 0. C. Doering III, Bell J. Econ. 7,287, (Spring 1976).
E. A. Copp and E. C. MacRae, “A Linear Programming Model of Emergency Oil Shortage Alternatives”, In Energy PoLcy
Evaluation (Edited by D. R. Lemage). Lexington Books, D. C. Heath, Lexington, Mass. (1974).
A. Ravindran and W. T. Begenyi, AIIE Trans. 8,258 (June 1976).
U.S. Department of Labor, BLS Handbook of Methods. U.S.Government Printing Office, Washington, DC. (1976).
U.S. Department of Energy, Altematiue Fuel Demand Due lo Natural Gas Curtailment, Washington, D. C., (1979).
8.‘0. Fisch, S. Gordon, W. Gorr, and B. v. Rabenau, “The Ohio Gas Allocation Model: Nonresidential Curtailments and
Employment Impacts”, In Modeling and Simulation (Edited by W. Vogt and M. Mickle), Pittsburgh (1979).
9. 0. Fisch, S. Gordon, W. Gorr, and 8. v. Rabenau, “The Ohio Gas Allocation Model: Weather Scenarios and Model
Simulation”, In Modeling and Simulation (Edited by W. Vogt and M. Mickle). Pittsburgh (1979).
IO. 0. Fisch, S. Gordon, W. Gorr, and B. v. Rabenau, “The Ohio Gas Allocation Model: Weather Scenarios and Model
Simulation”, In Modeling and Simulation (Edited by W. Vogt and M. Mickle). Pittsburgh (1979).

You might also like