I will let the data answer the question. To do that, I took the Cleveland Fed's monthly 10-year expected inflation rate series and plotted it against the

*subsequen*t growth in nominal consumer expenditures. I looked at a one, two, and three-year horizons for subsequent consumer spending growth. Below are the scatterplots created from this exercise. (The data starts in 1982:1 since that is the beginning of the Cleveland Fed's expected inflation series.)Not only is there a strong relationship, but it gets stronger at longer horizons for consumer expenditures. So changes in expected inflation do affect nominal spending. This is nothing new and is a central tenet of modern macroeconomics. I only bring it up now because some observers have questioned whether there really is this relationship. Reviewing this relationship also reminds us why it is important for the Fed to be clearer about the future path of monetary policy.

**Update:**I used expected inflation above because that has been the focus of recent debates. As an advocate of nominal GDP targeting, I believe a better perspective is to look at nominal spending expectations. To do that, I went to the Survey of Professional Forecasters and took the average annualized quarterly growth rate forecasted for nominal GDP over the next year and compared it to the actual growth of nominal GDP over the next year. Here is the figure:Once again, expectations matter.

David: any chance though, that you are just picking up the correlation between actual and expected inflation? In other words, we might expect to see some sort of positive correlation like this, even if real consumption was constant, and even if actual inflation was exogenous.

ReplyDeleteNick, just to be clear these figures are looking at expected inflation at time t and comparing it to actual nominal consumer spending at time t+1, t+2, etc.

ReplyDeleteIf changes in expected inflation at time t are causing subsequent changes in actual inflation at time t+1, then yes it is behind some of this relationship.

But that is the point, right? By having the central bank manage inflation expectations it really is managing what happens to actual inflation and other nominal variables. All I am trying to show here is that nominal spending can be influenced by expected inflation.

Is consumer spending on a per capita basis?

ReplyDeleteThe case for a casual relationship would seem stronger if spending normalized by nominal GDP growth increased faster when inflation expectations were higher. Or consider percent increases in real consumption on a per capita basis vs. inflation expectations.

Engineers like myself like to work the data hard.

Best regards

Kirk:

ReplyDeleteI went back took the and mid-month population(http://research.stlouisfed.org/fred2/graph/?g=1OM) and redid the graphs. The results were largely the same. For example, the R-squared was 46.14% for the 2-year horizon 52.42% for the 3-year.

David,

ReplyDeleteShouldn't expected inflation also cause the supply side in all markets like loanable funds and goods markets to shift to the left, in expectations of higher prices, therefore nullifying any positive shift from the demand curve? Why should expected inflation only effect demand but not also supply of loans, goods, and labor?

David, I initial reservation about the graph is same as Nick's. We know there is fairly high autocorrelation in both nominal private consumption and inflation. However, since this is market data and you are using 10y rather than short-term inflation expectations I think the problems are somewhat smaller.

ReplyDeleteHow would the graphs look in differences? So acceleration/deceleration in inflation expectations...

By the way it is somewhat surprising to me that Philly's show this low US inflation expectations...that of course just prove the point that the Fed needs to introduce a nominal target of some kind ASAP.

Lars,

ReplyDeleteI did this post in a rush and it is evident. I should have rotated the axis on the expected inflation graph so that x (expected inflation) causes y(consumer durables). I also should have done a vector autoregression and isolated true shocks to the forecasted series. Maybe I will do that in a note.