Excerpts from "Statistical Guessing Games",
The Economist, December 7 1996.
|
| A report by economists on the shortcomings of statistics
would not normally send politicians' pulses racing. But when it promises
to cut billions of dollars from the budget deficit, statistical detail
becomes big politics. A group of five eminent academics led by Michael
Boskin, a professor at Stanford University, has concluded that the CPI
probably exaggerates true increases in the cost of living by something
like 1.1% a year, although the exact figure could be 0.8% to 1.6%. |
| A correction of this magnitude would have enormous implications
for the budget. Around a third of federal spending, mostly in retirement
programs, is directly indexed to changes in consumer prices. Through the
indexing of individual income-tax brackets, a change in the CPI affects
federal revenues too. The Boskin Commission reckons that correcting the
1.1% overstatement would save the government around $1 trillion over the
next 12 years. By 2002, the target year for balancing the budget, around
$200 billion could have been saved. Small wonder, then that many politicians
hope to prune public spending by simply improving statistics. |
| What are the problems? The basic difficulty is that the
CPI is not really a measure of changes in the whole cost of living, but
rather a gauge of the price rises in a fixed basket of goods. This means
that the CPI does not keep up with changes in the products consumers buy,
or in how and where they buy them. According to the Boskin report, about
a third of the mismeasurement (0.4%) is due to the fact that the official
price index fails to capture important changes in consumer spending patterns.
If the price of a hardback book increases, for instance, people may buy
more paperbacks. A further 0.1% bias is due to the fact that Americans
now do more shopping at discount outlets. But the biggest effect (0.6%)
comes because the CPI underestimates the benefit shoppers gain from improvements
in product quality and from the plethora of new products. Cellular telephones,
for instance, are not yet included in the CPI (planned for Fall 1998),
although 40m Americans now own one. |
| The Boskin commission report makes a host of suggestions
as to how some of these inaccuracies could be corrected. But the CPI will
still overstate true increases in the cost of living. |
| This leads to the bigger political point. If the CPI,
however revised, is not an accurate guide to changes in the cost of living,
should it be the index to which retirement benefits are linked? The Boskin
report makes the point explicitly: Congress and the president, it suggests,
must decide whether they wish to continue widespread, substantial over-indexing
of benefits. In other words, if politicians are looking for budget savings,
this report gives them intellectual backing for proposing that retirement
benefits should be indexed to something less than the CPI. They could pass
a law, for instance, that set future benefit indexation at a rate lower
than the CPI, perhaps as part of a budget package. |