Chains and Baskets

A pound of whole-wheat bread today will run around $1.91; it cost $1.79 a year ago; and in 2001 the same loaf (all numbers are for May and all from the Bureau of Labor Statistics) ran $1.47. The first of those numbers was worth $0.14 in 1929—and would presumably also get you a loaf of bread. Welcome to inflation.

My subject today is how easy it is to track actual prices if you make the effort to collect them regularly, as the Consumer Price Index people do—and how complicated it gets to sum up all these numbers into a single CPI number by means of which we compare our current purchasing power to that of distant years—the marvel that in 1929 fourteen pennies bought us a loaf of bread, but in 2011 it too nearly 14 times more.

The BLS’s metaphor for this is to talk of a market basket of a curious kind. Everything goes into it, not just food and clothing but also such things as housing, medical care, transportation, education, communications, recreation, and everything else that we spend our money on. How exactly does the BLS determine what that shopping basket should contain?

It turns out that the market basket is defined by sampling. It involves 7,000 families who report every quarter on their spending habits. In addition, another 7,000 families maintain detailed diaries recording everything that they spent money on for a two-week period. All told, the shape and meaning of the market basket emerges every two years from 28,000 weekly diaries and 60,000 survey interviews. The diaries kept by these families provide the most important element of information of all, the relative weight that individual products or group of products should have in the determination of a single monthly CPI number. The families, of course, will report on food purchases, and in detail—as well as on their house payments or rents. Suppose food prices went up 5 percent but housing costs by only 1 percent. Which of these expenditures should be viewed as more important. The rules say that the more you spend, the more important that thing is. Therefore food will be less important than housing. Here, for instance, are the weights assigned to eight major categories in two different periods.

These weights come from those diaries. The 2005 values were applied in 2008 and subsequent years. The 2007-2008 values were applied in 2010 and subsequently. Here we note that some categories “lost” and others “gained” weight. Food lost some of its weight, therefore increases in food prices were reflected less and increases in recreational expenditures (the biggest upward change in weight) reflected more of their changes in the CPIs that used this kind of weighting. In other words, that loaf of bread at $1.91 reflected its 6.7 percent increase in price only partially. Only 14.91 percent of it (thus .999) actually managed to become part of the CPI.

Now, mind you, these are weights for aggregates. Transportation, for example, includes new vehicles, airline fares, gasoline, and auto insurance. Transportation dropped in “weight” in these two periods. Why? One would have to see the details. Did people buy fewer cars? Travel less by air? Did that more than match increases in gasoline prices? The devil is in the details. The basket contains more than 200 items—tangible like food and intangible like insurance payments. Each of these items has its own weight, and it is the aggregation of them that produces the category weights. Further, these are the weights used for Urban Wage Earners, and other categories are also collected separately and the later merged. Therefore things get complicated.

The actual job of obtaining the prices associated with the items in the basket is accomplished by in actual pricing-surveys conducted by the BLS itself.

Yesterday I mentioned a special deflationary index, the chained-dollar index used by the Bureau of Economic Analysis to express the value of Gross Domestic Product in constant dollars. It differs from the CPI in using the average of two successive periods (quarters, in this case), combining their weights and price increases, to give results for the current period. Of these two, lets call them a and b for the index to be used in Quarter A, b will be combined with c and averaged to produce the next chained index in Quarter B.

Chained, ah, baskets seem to be a little more representative of real change than just baskets with their weights changed every now and then, say at two-year intervals, rather than quarterly, with two quarters’ results averaged or smoothed out.

Seven thousand people keeping diaries in the midst of life’s hurly-burly. Then mountainous computer work and 60,000 interviews just to define the basket. Finally, after the actual price survey, comes the precious number or its chained cousin. But by the time the number hits the news, it has the sound of Gospel.