About Bubble Tracker

CPI is a well-constructed index for its intended purpose: measuring the cost of a representative consumption basket for urban households. But every price index embeds choices about scope (what goes in the basket), weighting (geometric vs. arithmetic means), quality adjustment (hedonic regression, substitution effects), and coverage (CPI excludes investment assets by design; PCE uses a different weighting methodology and broader scope). These are defensible choices, not flaws, but they mean CPI captures one particular slice of the price landscape. Understanding what CPI measures, and what it does not, is the starting point for reading this site.

Monetary policy operates through multiple channels simultaneously: short-term rates, long-term yields, term premia, bank balance sheets, credit conditions, portfolio rebalancing, and expectations. The transmission speeds differ: financial asset prices can respond within hours; housing markets adjust over quarters; official shelter CPI, which relies on lagged rent surveys and imputed owner-equivalent rent, takes 12-18 months to reflect market conditions. Bubble Tracker puts these different price domains on the same timeline so users can inspect co-movement, divergence, and lag structure for themselves.

We do not claim CPI understates 'true' inflation, nor do we propose a single alternative number. Instead, we present multiple price series transparently: consumer prices, asset prices, and the monetary aggregates that may link them, alongside the statistical uncertainty inherent in each relationship.

Richard Cantillon observed in the 1700s that whoever is closest to the source of new money benefits most. The modern central-bank literature on distributional effects is more nuanced: easier policy can lift asset values and benefit households with larger asset holdings, but employment and income channels can partially offset that effect.

We estimate reduced-form elasticities of each asset class to M2 money supply. In our baseline sample, equity prices show a substantially higher estimated elasticity than CPI. This is consistent with the idea that financial markets respond more quickly to monetary conditions than consumer prices.

These estimates should be read as reduced-form associations, not as a literal accounting of where every new dollar went. Trending macro series can generate misleadingly high R² and t-statistics if unit roots and cointegration are not handled carefully, which is why we report bootstrap intervals, cross-validation, rolling windows, and out-of-sample backtests alongside point estimates.

Whether markets are “overvalued“ depends on both the lens and the horizon. Different ratios ask different questions: P/B measures how far price stands above recorded equity; Forward P/E asks how dearly investors are paying for expected near-term profits; CAPE compares price with a long-run, cycle-smoothed notion of earnings power.

Equity valuations move with expected cash flows, discount rates, and risk premia. A constant-growth sketch: P/E ≈ 1 ÷ (r − g). When real yields fall, (r − g) shrinks and the same profit stream is worth more today. If multiples rise without EPS upgrades, valuation is being driven by rates rather than fundamentals.

Campbell-Shiller's work supports valuation ratios as useful for multi-year return outlooks, while short-run returns are dominated by flows, risk appetite, and macro shocks.

• P/B is sensitive to sector mix and intangible capital, which is only partly reflected on balance sheets.
• Forward P/E reflects both earnings expectations and discount rates; multiple expansion without EPS upgrades is a sign the advance is rate-driven.
• CAPE smooths the cycle and aligns better with long-horizon returns than with short-term timing.

The word 'inflation' is used loosely in public discourse but refers to several distinct phenomena. CPI inflation measures changes in the cost of a consumption basket: it is a cost-of-living concept. Asset-price inflation (rising equity valuations, home prices, bond prices) reflects changes in discount rates, risk premia, and expected cash flows: it is a wealth concept. These can move in opposite directions: a period of low CPI inflation with rising asset prices is not a contradiction; it is two different channels of monetary transmission operating at different speeds.

CPI itself involves non-trivial methodological choices. Hedonic quality adjustment (e.g., for electronics, vehicles) attempts to separate price changes from quality changes, reasonable in principle but consequential in practice. The substitution effect (consumers switching to cheaper alternatives when prices rise) means the Laspeyres-type CPI can overstate cost-of-living inflation, while the BLS's geometric mean formula partially corrects this. Owner's Equivalent Rent, roughly 25% of CPI, imputes housing costs for homeowners using rental survey data rather than actual home prices or mortgage costs, creating a significant measurement lag and a conceptual gap between lived experience and the index.

PCE (Personal Consumption Expenditures), the Fed's preferred measure, uses chain-weighted Fisher indices and broader coverage including employer-provided health insurance. PCE typically runs 0.3-0.5 percentage points below CPI. Neither is 'right'; they answer slightly different questions. We primarily use CPI because it has longer history and monthly frequency, but the choice of index matters for interpretation.

• Lower real rates → higher present values → richer multiples. This is a valuation channel, not 'inflation' in the CPI sense, but it affects wealth and purchasing power for asset holders.
• OER methodology means CPI shelter lags market rents by 12-18 months, a measurement artifact, not an economic mystery. This lag is one of the most stable empirical regularities we track.
• M2–CPI correlations are regime-dependent: strong during high-inflation episodes (2021–23), weak in the low-inflation 2010s. The BIS (2023) confirms that money growth carried useful information post-pandemic but was much less informative in the prior decade.

The site is designed to be explored, not just read. Every chart is interactive: zoom, pan, and compare. Charts on the same page sync their x-axis, so zooming one zooms them all.