The Inflated Case Against the CPI

There is now the appearance of an expert consensus that the government's most important measure of inflation, the consumer price index (CPI), seriously overstates the true increase in the cost of living. This sudden enlightenment is less the result of new research than political convenience. A cut in the CPI would reduce government payouts and ease the path to deficit reduction. Even better, it would do so via a technical adjustment that left few political fingerprints.

Tax brackets and government benefit programs such as Social Security are indexed to the CPI. If the CPI overstates inflation by 1 percent, as the Senate Finance Committee's Boskin panel has proposed, and the index is adjusted accordingly, this would reduce benefits and the deficit by a cumulative total of $634 billion over 10 years. Not bad for a technical fix.

Doubtless, the way we measure inflation requires continuous refinement. The Bureau of Labor Statistics (BLS) takes this task seriously, and has made myriad small adjustments over the past three decades. For years, there was a nuanced and relatively obscure debate about how to fine-tune the CPI. Lately, there has been a politically driven frenzy, as a small group of economists has labored to uncover--and exaggerate--all the ways in which the CPI might overstate inflation. Many claims have been advanced based on very little real evidence. There has also been virtually no effort to examine the ways in which the CPI might understate inflation.

The immediate protagonists are the five-member panel appointed by the Senate Finance Committee, chaired by former Bush economic advisor Michael Boskin, to make recommendations on revisions in the CPI. Though the group includes some eminent economists, all had previously testified on the CPI's supposed bias. All were chosen as known quantities who could be reliably counted upon to recommend a downward revision. Other eminent economists such as former BLS Commissioner Janet Norwood, who took the opposite view, were ignored. The panel was appointed in June 1995 and announced its 1 percent solution in mid-September. It conducted no original research. Instead, it used rough rules of thumb to reach its conclusions.




Five factors are usually cited by those claiming an upward bias in the CPI. Each provides some basis for claiming the index overstates inflation. However, the size of any resulting overstatement is far smaller than what is being claimed, and may well be offset by the sources of understatement in the CPI.

Substitution Effects. This is the most frequently cited source of bias, perhaps because so many reporters learned about it in their introductory economics classes. Most goods have close substitutes. If oranges are $1.99 a pound, consumers switch to apples. The CPI measures the prices of a fixed basket of goods and services. When the price of some goods in this basket rises temporarily, thrifty consumers shift to substitutes. By holding the basket fixed, the CPI then overstates the true increase in the cost of living for most consumers. This is a fair criticism.

However, most studies that have tried to measure the size of this bias find it to be very small, between 0.1 percent and 0.2 percent annually. (The Boskin panel scored it as 0.3 percent.) Moreover, one might fairly argue that even a close substitution "choice" dictated by a price rise entails an offsetting loss to quality. Presuming the substitute to be identical is like comparing, well, apples and oranges.

The Wal-Mart Effect. Over the last several decades discount stores have displaced many traditional retailers. As a result, consumers purchase many goods at far lower prices. The CPI does base its local samples on where consumers actually shop, but the CPI does not record the switch from a traditional department store to a discount store as a price decline. Rather, the price differential is treated as offsetting the lower quality of service in the discount store. Clearly this treatment misses a cost saving. Since discount stores have grown rapidly at the expense of traditional retailers, many consumers must consider the cheaper price well worth the lower quality service.

However, the importance of this difference for the CPI has been vastly overstated. Only about 15 percent of the index consists of goods that could potentially be sold in discount stores (primarily apparel, appliances, and household furniture). The share of consumers who patronize discounters versus full-price retailers simply does not change much from year to year. Moreover, even if the true price difference is as much as 10 percent (after adjusting for differences in service quality), this would lead to a bias of just 0.015 percent a year. This compares to a figure of 0.2 percent to 0.4 percent often cited by those claiming a substantial CPI overstatement of inflation. The Boskin panel used a figure of 0.2 percent.

Subscribe to The American Prospect

Quality Bias. One of the largest sources of alleged overstatement of inflation is the failure of the consumer price index to fully account for the improvements in product quality. Clearly, a $2,000 computer today is a far superior machine to a $2,000 computer bought as recently as 1994. Most products are continuously improving in quality. But in fact, the consumer price index already includes extensive adjustments for product quality. It may even overstate the improvement in quality in some cases.

For example, the price index for new cars has increased approximately 150 percent from 1970 to the present, although the average price of an actual new car has increased about fourfold. Based on the CPI's price adjustment, it should be possible to purchase a car today for approximately 2.5 times what a comparable car cost in 1970. In 1970, a new Volkswagen Beetle cost about $2,000. But no new car is on the market today for anything close to $5,000, as would be implied by the new car index. The bottom-of-the-line new car available today for $9,000 is doubtless a significantly better car than a Volkswagen Beetle, but there are probably many consumers who would prefer to purchase a new Beetle at $5,000 rather than pay more for a better car.

This is the general problem with the quality adjustments in the index. Price increases attributed to quality adjustments are not counted as price increases in the index, even though many consumers might not pay for them if they had the choice.

There are also many areas where quality has plainly deteriorated. For example, many consumers have to spend more time fighting with health insurance companies over the processing of claims than 20 years ago, but this deterioration is not picked up in the index. Airline service has declined--less legroom, more changes of planes and missed connections, fewer meals, more convoluted fares. Other examples include longer waits in traffic and time spent waiting on hold or navigating voicemail instructions. None of these quality deteriorations are recorded in the CPI. Given the very limited research on quality adjustments in the CPI (the most frequently cited work is now twelve years out of date), it is impossible to reach any conclusive judgment about the size, or direction, of quality bias in the present index.

New Products. New goods are typically not included in the index until several years after they first appear on the market. During this time, they often undergo large price reductions--which are not picked up in the index. The inclusion of these price declines would lead to a lower measure of inflation. The classic example of this problem is the hand calculator. When it first appeared on the market it cost over $1,000. Within a couple of years its price had fallen to under $100. This huge decline in price was not picked up in the CPI.

But wait. Most people don't buy very expensive new products. Such products only find mass markets after their price has come down dramatically. (How many people bought thousand-dollar hand calculators?)

The CPI is an "expenditure- weighted index," meaning that each dollar of consumer expenditure is weighted equally. This means that if Donald Trump spends 1,000 times what the average consumer spends, his expenditures count 1,000 times as much as those of the average consumer. Of course, it is the Donald Trumps who buy the expensive new products.

However, it is possible to construct the index in a different way, which would count each consumer's expenditures equally. Under this system 1 percent of my budget would count the same as 1 percent of Donald Trump's budget in determining the weighting of a particular item. Such an approach is clearly more appropriate for the purposes the CPI is intended for. There is no reason that Social Security recipients should get a smaller cost of living adjustment because a few wealthy people experience huge savings on hand calculators. Nor does it make sense that such a bonanza for the wealthy should affect wage contracts that use the CPI as a point of reference.

A "person-weighted" index that counted each individual's expenditures equally would provide a much better gauge of the increase in the cost of living experienced by most of the population--and it would virtually eliminate the problem of new goods as a source of bias in the index. Here is a case crying out for a genuine technical adjustment, but one that cuts in the opposite direction from the Boskin panel.

Formula Bias. This is the most technical issue (sorry). Suppose the price of a good rises from $1.00 to $1.10. This is a 10 percent increase in price. Suppose it then falls back from $1.10 to $1.00; this is a 9 percent decrease in price. If these changes were just added together it would imply a 1 percent increase in price even though the price had not changed at all. BLS was never so foolish as to construct the CPI in such a way that it would be generally subject to this bias. But there are other such technical problems. BLS has researched this issue extensively, and uncovered and corrected several such areas. It is possible that problems of this sort may still exist in places, but the impact is likely to be extremely small, almost certainly less than 0.1 percent annually.

Let's give the critics the benefit of the doubt. Adding all of these possible biases together, a plausible estimate of inflation overstatement in the CPI would be 0.4 percent--not the one full percentage point estimated by the Boskin panel. But this is a gross adjustment, not a net one: It revises only those elements of the CPI that apparently overstate inflation. To be accurate, one needs to offset this adjustment with factors that suggest the CPI may be too low.



The possibility that the CPI understates inflation has received little attention, since it won't help cut the deficit. Indeed, if the CPI errs on the low side, then Social Security pensioners should be getting bigger checks. Consider three distinct sources of downward bias:

Health Insurance Costs. The CPI does not include most increases in insurance premiums for individual health insurance, or increases in copayments or deductibles on employer-purchased insurance. This has been a major drain on household budgets in recent years, as employers have shifted more of the cost of health insurance back to their employees. Nor does the CPI include increases in required payments by beneficiaries of government programs such as the proposed increases in the Medicare Part B premium. This exclusion would lead to an enormous understatement in the cost of living of an elderly household. Even under the Clinton Medicare proposal, the premium is scheduled to rise by approximately $500 per beneficiary over the next seven years. This increase will consume nearly 5 percent of the annual income of a couple with an income of $20,000, the midpoint of the income distribution for families over 65. Under the Gingrich proposal, the total increased costs to consumers would be substantially higher.

Today's basket of health services is in many respects superior to that of, say, two decades ago--thanks to new technologies, drugs, and lifesaving procedures. It's also true that under managed care, many doctors are more harried and patients are often rushed out of hospitals. Some of the higher cost of today's health care reflects not better service, but deadweight losses--the cost of claims processing, risk selection, mergers, and acquisitions. Spending on health care has quintupled in three decades, but it's not at all clear that the quality has. It would take a great deal more research to determine how to net out the improvements and degradations to quality, and then weigh them against the unambiguously higher cost.

Personal Business Expenditures. This category of expenditures, which includes items such as lawyers' and brokerage fees, has been rising at the rate of 0.2 percent a year for the past 20 years as a share of disposable income. This is a category of spending that may not provide direct benefits to consumers but rather is often a cost of maintaining a standard of living threatened by deteriorating external circumstances. Suppose, for example, that I have to hire a lawyer in order to resolve a dispute with my health insurance company, because health insurers are becoming more aggressive. Compared to a situation where the insurance company deals with me honorably, this is a needless expense. I am worse off in direct proportion to the amount of money that I have to spend on my lawyer, regardless of the quantity or quality of legal services provided.

Similarly, if I have to rely on a private financial counselor to ensure a secure retirement, instead of receiving a company pension, I am worse off in direct proportion to the amount of money I must pay the counselor, not the fee per financial transaction. Likewise the cost of divorce lawyers and security consultants. Since the CPI only measures the change in the price of these services, rather than the change in the overall need for the services or genuine benefit derived, it is understating the true increase in the cost of living.

Quality of Life Factors. Many factors that affect the quality of life are not picked up in the CPI. An obvious example is crime. If people have to spend more money in order to live in a safe neighborhood--say on security alarms--it is not captured by the CPI. Nor is a deterioration in the quality of public schools and therefore an increased need for spending on private schools or personal tutors. Nor is the cost of joining a private exercise club because the local public facility has closed or become unsafe. Many of these issues are quite complex and cannot be easily quantified in any meaningful way, but this doesn't make them any less real. Anyone attempting to make conclusive statements about changes in the "true" cost of living must be prepared to address such issues.

Given the very limited amount of research on the biases in the CPI, it is impossible at this point to reach any conclusive judgment about the magnitude or direction of the overall bias in the index. While the claim that "the CPI overstates inflation" has become a virtual mantra of the Washington punditry, honest proponents of this view acknowledge that it is based on very little evidence. The best solution would be to provide the professional statisticians at BLS with the resources they need to improve the CPI. This is clearly preferable to bending economic statistics in whatever direction is politically expedient.


Unexpected Twists

One striking thing about this whole debate is that conservatives try to have the argument both ways. Supposedly, a lower CPI would cut the deficit; and cutting the deficit would stop us from "robbing our grandchildren." But if the Boskin panel is right and the CPI is overstated by 1 percent per year, then real median wages are rising at 2 percent per year. This means that they will double in approximately 35 years, so that the average real wage will be approximately $50,000 a year (in 1995 dollars) in the year 2030. If so, our grandchildren will do just fine and there's no need to slash the Social Security of their grandparents.

An overstated CPI also means that people had been much poorer in the recent past than we realized. If the CPI was overstated by 1.5 percent in the past, as suggested by the Boskin Commission, then the average annual wage in 1960 was just $11,215 measured in 1995 dollars. In 1960, today's 70-year-olds were 35. It is hard to justify taking Social Security benefits from these people in order to make the 35-year-olds of 2030 better off.

In addition, the critics have ignored the implications of a significant CPI revision for monetary policy and growth. Alan Greenspan, chairman of the Federal Reserve, has testified that the CPI overstates inflation by as much as 1.5 percent. But if the official inflation rate is lowered, it undermines the entire rationale of Greenspan's tight money policy. So politicizing the CPI leads to some unexpected implications. It would be far better to return the question of CPI revision to intellectually honest technicians, where it belongs, and to argue deficit reduction on its merits.

You may also like