© 2000 John Petroff
|
|
© 2000 John Petroff |
| No rated * * * * * | Resize -A +A |
Sensitivity analysis is used to predict the most likely changes in demand for products of an industry (or company) given some known trends in variables to which the industry is most sensitive. The mathematical aspects of the technique have been described in Chapter 5 Section E and their use in Chapter 9 Section E-2. Some of the major long term trends and short term attitude variables that are commonly studied will be discussed in Chapter 15 Section A. In this chapter, we show how these variables are integrated into an industry analysis.
The method consists in using historical data of industry sales
and a set of variables listed below to identify which of these
has the most impact on industry sales. Least-squares
regressions are run on as many years as possible to determine
which variables are the most correlated with changes of industry
sales. Some of the major economic variables
are
- total population, number of births, number of deaths,
- population break-down by age, gender, marital status, education,
- gross domestic product, national income, consumption expenditure,
housing starts,
- gross and net domestic investment, output of durable goods,
- rate of saving, stock market indexes,
- consumer price index, wholesale price index, average hourly
salary,
- unemployment rate, length of work week, proportion of temporary
workers,
- interest rates on government securities, commercial bank prime
rates, mortgage rates,
- regional data for all of the above.
Often economic variables are not
stated in absolute current terms, but in real rate of change.
Thus some of the most common economic variables to be tried with
the industry rate of sales growth are
- annual rate of change in national income,
- population growth rate,
- rate of growth of gross domestic product,
- inflation rate.
The exogenous variables that should be retained from the above
list of variables, are those that have the highest explanatory
power (i.e. in statistical terms, that would mean a high t statistic
of at least 2, with an overall R2 higher than .50,
as explained in Chapter 5 Section
E-1), and that can be reasonably assumed to have a causal
effect on the dependent variable (i.e. industry sales).
Combined with economic variables are usually consumer attitude variables. These are gathered from surveys conducted by the Conference Board and the University of Michigan. For industries affected by international trade, annual volume of exports and/or imports, balance on current account and exchange rates would be used. For manufacturing, plant capacity utilization rate is important. For financial sector industries, monetary aggregates M1, M2 and M3, as well as yields on various debt instruments are some of the variables used. For many industries, costs of materials would be an important exogenous variable. This would be true for fuel for airlines, lumber for construction, electricity for aluminum, etc. In addition, the impact of proposed legislation can be imbedded in the regression with some quantitative proxy.
After obtaining robust regression results that explains the historical pattern of industry sales, the analyst must find the most recent projection of the explanatory variables for the coming year or phase of business cycle. Occasionally, the sensitivity analysis is extended beyond the current business cycle; but this is rare because economic trend projections beyond two or three years is far more chancy and attitude variables can no longer be used. The last step is the easiest: calculate the value of the endogenous variable.
For example, an analyst wants to forecast the change in new housing starts for next year in a given region. His regression analysis for the past 20 years in the region has shown that the rate of growth of housing starts is explained in equal parts (i.e. 33%) by population growth, drop in mortgage rates and drop in price of lumber. If the economic and demographic patterns remain the same the following year, but mortgage rates are expected to rise by one percent, the analyst can predict that the rate of growth housing starts in the region will drop by 0.33% next year.
See review questions Q-14E.1 through Q-14E.7.
See research assignment R-14E.1.
| Previous: Cyclicality |
|
Next: Technology |