The Model
It is possible to make a !oore definite analysis of the
effect of jawboning and pr ice controls. As indicated earlier,
a cruc ial test of the impact of these Government act ions on
the steel industry is whether they reduced steel profits.
Therefore, a measure of the determinants of profits in the
U.S. steel industry can be used to conduct a statistical test
of whether jawboning or price controls, or both, adversely
af fected the industry.
The time per iod used for th is test was 1961 through 1975,
beginning with the third quarter of 1961 when the Kennedy
Administration made its first attempt to prevent a steel price
increase. All var iables were measured on a quar ter ly basis.
The dependent var iable, steel industry net prof i ts after taxes,
was deflated by the Bureau of Labor Statistics wholesale price
index for all commodities in order to measure steel profits
i
,
more nearìy in terms ~f purchasing power than would be the
-257-
case without such adjustment. The source of the profit data
was the Quarterly Financial Report for Manufacturing Corporations issued by the Federal Trade Commission.
The independent variables selected were raw steel production, steel imports, and unfilled orders in the steel industry,
and to these were added dummy var iables for jawboning and pr ice
controls. (The source for production and for steel imports was,
.,.,.'.
-.
the Annual Statistical Report of the Amer ican Iron and Steel
Institute, and the source for unfilled orders was the Commerce
Department's Survey of Current Business.) Unfilled orders wasused as an indicator of price changes not reported in the BLS
pr ice indexes for steel, such as discounts, changes in extr a
charges, freight absorption, warehouse charges, and cred it
terms. (Since the Commerce Department reports unfilled orders
in dollars but not in tons, this series was deflated by the
ratio of steel shipments in dollars to steel shipments in
tons. )
The dummy variable for the price controls of 1971-74 was
a value of one assigned to each quarter in which the controls
were in effect and a value of zero assigned to all other
quarters.
A single-stage ordinary least squares model was first used
to test these hypotheses. Eight different dummy variable
systems were tried, but none indicated a significant relationship between steel profits and jawboning. However, all these,
results yielded a posi~ive relationship between jawboning and
-258-
steel profits. Since there is no justification for a hypothesis
that jawboning increases steel profits, there was reason to
believe that the linear model used for the first set of tests
was biased. Such bias can be caused by a correlation between
the explanatory variables and the error term which is normally
present in a s imul taneous equat ion model. In th is case, it
seems reasonable to assume that the profit equation is part
. r. .
of a larger simultaneous system. This simultaneous system
also describes the process which determines the decision concerning when to jawbone steel price increases. And if some
of the same var iables affect both steel prof i ts and the j awboning decision process, a bias would be expected in the
ordinary least squares estimates. This source of bias can be
overcome by a procedure that explains the systematic part of
the jawboning decision which is not explained by the variables
included in the error term of the profit equation. l/ This
procedure took the form of a two-step analysis: One equation
was used to estimate the probability of jawboning during each
three-month per iod, and the estimated probability of jawboning
was used as an instrumental var iable in a second equation to
estimate the relationship between steel profits and jawboning.
The hypothesis underlying the first equation was that
Government officials would be more likely to jawbone steel
