While cross-correlations are instructive, they cannot fully capture complicated dynamics. To assess predictive performance, we first use in-sample or within-sample regressions.
Predictions based on in-sample estimation uses the entire data sample to estimate the parameters in order to best predict observations within the sample. We focus on the ability of each inflation expectations measure to predict one-year-ahead inflation. The specification we use is. We estimate our specification over two sample periods: January through May and—to accommodate the shorter availability of the Atlanta Fed survey—October through May We first report results from a baseline specification that excludes any inflation expectation series.
The change in the regression R 2 as we move from this baseline specification to one that includes a particular inflation expectation variable tells us how much that variable contributes to improved prediction, that is, how much it reduces the prediction error. Coefficient estimates that appear in bold are statistically significant at the 5 percent level. If the inflation expectations term in a given regression is a useful predictor, it will increase R 2 notably.
The main takeaway from this exercise is that Blue Chip has consistently been the best inflation predictor, featuring the highest R 2 and regression coefficients that have often been closest to 1, suggesting a tighter relationship with actual inflation, while the Michigan consumer survey has been the worst predictor. The assessment and ranking of the other two measures are more complicated. The Cleveland Fed model performs well over the full sample period but is a rather poor predictor over the period from onward.
Over this recent period, the Atlanta Fed business survey is usually the second-best predictor, although its performance still leaves much to be desired. Out-of-sample forecasting does not use the entire data sample at once. Instead, for a forecast of month-ahead inflation at date t , it uses information in the sample only up to date t.
While the in-sample regressions discussed above suggest that inflation expectations may have some predictive power for inflation over the data sample, out-of-sample forecasting is often a more difficult challenge because forecasts are based only on data available at the time the forecasts are produced, and therefore relatively weak relationships may evolve or simply break down over time. In this vein, we conduct a forecasting exercise in which we treat each inflation expectations estimate at month t as the month-ahead forecast of CPI inflation, and we compare this prediction to the realized or actual inflation reading one year in the future.
We also compare the accuracy of these CPI inflation forecasts to a CPI inflation forecast produced using the popular benchmark forecasting model of Stock and Watson since recent research has documented that such time-series models can produce forecasts that rival those of surveys of professional forecasters. Our chief interest is in predicting CPI inflation.
However, we also evaluate the extent to which these inflation expectations measures forecast future core CPI inflation, median CPI inflation, and trimmed-mean CPI inflation. We determine whose inflation expectations provide the best signal about the inflation outlook. In the longer evaluation sample, Blue Chip outperforms the Michigan survey and the Cleveland Fed model for each of our inflation measures at the one-year horizon.
In the shorter evaluation sample that includes the Atlanta Fed survey, Blue Chip and the Atlanta Fed survey outperform the other two measures. Across the board, the Michigan survey fares poorly compared with the other inflation expectations, while the Cleveland Fed model is in the middle. However, information about the inflation trend has been compared to a radio signal that is obscured by static.
Just as noise filters are used to remove the static in radio signals, economists filter inflation data to remove the static caused by supply and demand changes. One way to filter the inflation news is to measure the change in prices over a long period, such as a year, to eliminate the short-run fluctuations. But then, the useful information is delayed for a year. Another way that economists filter out the static is to delete the items in the price index that are sensitive to large, frequent disturbances to supply and demand and, therefore, have highly volatile prices.
After deleting these items, what is left is core inflation , that is, inflation in the basket of goods excluding the more volatile components. Since the s, core inflation has typically been measured by excluding food and energy from the basket of goods. This is because the early s saw highly volatile food prices and, soon afterward, a rapid rise in the prices of gas, oil and other energy products.
The core measure of inflation, the PCEPI excluding food and energy, has been less sensitive to temporary shocks to the economy and has seemed to have been a better barometer of the underlying trend in inflation than the all-item PCEPI. Looking at Figure 1 , we see that the rate of inflation measured by the PCEPI excluding food and energy has been less volatile than with the all-item index.
When inflation dropped considerably in the middle of , the index excluding food and energy did not show the same massive drop. Let's take a closer look at the changes in the prices of components excluded from the core: food and energy. From Figure 2 , we see that inflation in energy prices indeed has been very volatile, increasing and decreasing much more than the food component or the all-item PCEPI.
We also see that food prices have become increasingly stable recently, while energy prices continue to fluctuate significantly. What has caused the recent increase in the stability of food prices? Improvements in technology and a change in consumer eating habits have both contributed. It is not unusual, as it once was, for a shopper in a supermarket in Chicago to be buying fresh produce grown in South America. As technological advances have reduced the cost of air freight and refrigeration, their use has become widespread and commonplace in the food industry, increasing the geographic size of the market for food and reducing the volatility of food prices.
Another change in the food distribution system is that many more people now buy their food from large grocery store chains. These large chains have an advantage over smaller specialty retailers in that they have the ability to stock larger quantities of many more different types of items.
Large supermarkets purchase food directly from the producers in huge quantities, cutting the cost to themselves and their consumers. Eating habits of the American consumer also have changed. With the hectic schedule many Americans have, people are less inclined to buy fresh fruit, vegetables, meat and poultry that may go bad in their refrigerators or require time and energy to prepare.
People are much more likely to buy prepared meals at the grocery store or to eat at restaurants. The prices that consumers pay for these meals are largely expenditures on the labor used to prepare and serve the food. The price of these labor services is less volatile than is the price of the raw food products. Inflation forecast is measured in terms of the consumer price index CPI or harmonised index of consumer prices HICP for euro area countries, the euro area aggregate and the United Kingdom.
Inflation measures the general evolution of prices. It is defined as the change in the prices of a basket of goods and services that are typically purchased by households. Projections are based on an assessment of the economic climate in individual countries and the world economy, using a combination of model-based analyses and expert judgement. The indicator is expressed in annual growth rates. Germany , G20 12th Summit Are you sure you want to delete this page? Are you sure you want to delete this document?
Unable to delete the page because it has shortcuts referencing it at the following locations: Please delete these shortcuts first, then delete page itself. Knoema, an Eldridge business, is the most comprehensive source of global decision-making data in the world. Legal Terms of Use Privacy Policy. Newsletter subscription You're subscribed! Please provide valid e-mail Subscribe.
0コメント