Madness: Economics, the Federal Reserve, inflation, interest rates, and damned statistics

"Brookings Economic Studies just posted an article about inflation rate reporting. The President doesn't like high inflation rates: they make him look bad so he kills the messenger, Jerome Powell, and asks for new standards. His new pick, Kevin Warsh, says he wants to use 'trimmed averages.' This sounds like a shell game that might appease the president, but not really do anything.

You can look at various tools for appraising inflation. There is the mathematical average (the mean), the mathematical most frequently appearing or most common result (mode), and the trimmed average that knocks off outliers on both sides.

When I do statistics I knock off the outliers because they likely don't reflect the most common result. They just throw it off. For example, early deaths in the first 50 years of life drag down the average life expectancy figures (79 years) so they don't show how long the typical life expectancy is (mode: 87-90 years). So the average scares a lot of people unnecessarily. To get around this, either look at the mode or trim the outliers.

The more accurate tale is presented by a rolling average. To get a rolling average, take the mode from the last 3 months and average it. Next month do the same. This gives you a much more reliable statistical indicator with the effect of noise, endless variability such as holidays, and outliers. It's a much better indicator of whether interest rates need to be changed or social security payouts should be hiked."

Discussion: Political Optics vs. Statistical Reality

This critique highlights the fundamental tension between a metric that captures the "whole truth" versus one that captures the "persistent trend." While the Trimmed Mean CPI is used by the Cleveland and Dallas Feds to solve for volatility, the timing of such a change often signals a "shell game."

• The Mean (average): Sensitive to "noise." A single category spike (like eggs) can trigger unnecessary panic.
• The Trimmed Mean: By cutting the "tails" of the distribution, it focuses on inflationary inertia. However, choosing a specific trimming percentage to hit a political target undermines the Fed's credibility.
• The Mode (the mathematical typical or most frequent): Reflects the "typical" experience. Much like life expectancy figures, where the mean (79) is dragged down by early deaths, the mode (87-90) shows the most common reality.

The Power of Rolling Averages

A rolling average of the mode is a sophisticated way to filter out seasonal noise (holidays, summer travel). It prevents knee-jerk reactions, ensuring policy shifts like interest rate adjustments and Social Security hikes are based on a "heartbeat" rather than a temporary "palpitation."

Quantifying the "Spiral": Solving the Rolling Average Lag

"The more you 'smooth' or 'roll' the data, the slower you are to see a genuine new trend starting. If inflation is truly spiraling, a rolling average might keep the Fed from acting until it's too late. A subsection of this might be trend watching that separates out trends to watch and quantifies them. This requires some refinement."

Time Series Decomposition

To prevent the "smoothing" from becoming a blind spot that hides true trends, we can break data into three components:

1. The Secular Trend (Long Game): Strips away everything to see if the dollar is fundamentally eroding over 5–10 years.
2. The Cyclical Component (The Spiral): This is the "danger zone." By subtracting the trend from the rolling average, we find the "Cyclical Residual." If this grows for three quarters, the spiral is real.
3. The Stochastic Component (Noise): The random spikes (outliers) that should be trimmed to find the true signal.

Using Velocity vs. Acceleration to understand trends

To quantify these for a "60-second read," we can borrow from systems engineering:

• Velocity: The 3-month rolling average of the Mode (Where we are).
• Acceleration: The change in that velocity (How fast we are changing).

If Velocity is 3% but Acceleration is positive, the "spiral" is active, and the rolling average lag is no longer a safety buffer—it's a warning.

Or a simpler method is to identify sharp rises in numbers that could become trends and isolate them into a watch category to see if they continue spiraling.