Evaluating headline versus food-and-energy-excluded CPI as bond-yield context

Two inflation series as yield context

A long-bond yield move can be placed in intermarket context by pairing it with a consumer inflation series. Two candidates compete: headline CPI, the all-items consumer inflation rate measured here as the change over the prior twelve months, and food-and-energy-excluded CPI, the consumer inflation rate that omits food and energy prices and is often treated as a smoother alternative to the all-items series.

The historical workflow compared each series with the 30-year yield. It did not treat the smoother series as automatically the better backdrop.

First pass: rolling co-movement

A correlation coefficient is a signed association statistic between two series. It runs from -1.0 to +1.0, and zero implies no linear relationship. It records both the direction and the strength of association.

Monthly data covering 16 years were examined in overlapping two-year windows advanced one year at a time. That rolling two-year window design produced 15 coefficients for each pairing of the 30-year yield with an inflation series. Successive windows overlap, so the 15 coefficients can be averaged.

Across those 15 windows the average coefficient was +0.50 for yield versus headline CPI and +0.31 for yield versus food-and-energy-excluded CPI. The yield versus headline CPI coefficient exceeded the yield versus food-and-energy-excluded coefficient in 10 of the 15 windows.

Second pass: channel-defined yield trends

A moving-average channel was then used to restrict the comparison to lengthy yield trends. The channel was a band around a 10-month average of yield, with the bands set at 2.5 percent of that average above and below it. Crossing the channel left behind a reactionary turn: the prior high or low used as the start or end of a trend segment.

A yield-trend segment is a lengthy up-move from a reactionary low to a reactionary high, or a down-move from a high to a low. The channel marked reactionary yield highs and lows and isolated those lengthy trends.

Across 14 channel-defined yield trends, yield change and headline CPI change shared direction in 10 cases, versus 7 cases for the food-and-energy-excluded series. Over those 14 trend segments the correlation of yield change with headline CPI change was +0.79, versus +0.53 with food-and-energy-excluded CPI.

One uptrend as a check

In the March 1986 to September 1987 yield uptrend, yield rose 2.33 while headline CPI rose 2.21 and food-and-energy-excluded CPI rose 0.12.

Editorial: the first pass answers which series co-moves more often in overlapping two-year samples. The second pass asks whether that comparison still holds when the sample is only the yield-trend segments that the channel isolates. On both passes the historical coefficients were larger for headline CPI. The March 1986 to September 1987 uptrend is one case in which yield and headline CPI rose together while food-and-energy-excluded CPI rose 0.12. This is a historical comparison of context series, not a claim about present-day yields.