Skip to main content

Why Headlines Hide Real Volatility by Choosing the Window

A headline announces: "Unemployment Falls for Fifth Consecutive Month." It sounds like solid improvement. But a more honest headline might read: "Unemployment Falls for Fifth Month, But Up 40% from Year Ago." The same data. Completely opposite stories.

Or: "Stocks Post Best Quarter in Three Years." Impressive. But what if this quarter's gain only recovered losses from the previous two quarters? The headline focuses on the recent window (positive) while ignoring the broader context (flat or negative overall).

This is the rolling-average trap. Financial news reports data over arbitrary windows—the last month, the last quarter, the last year—and the choice of window dramatically changes whether the story is "things are improving" or "things are getting worse." Journalists choose the window that makes the story most dramatic or that aligns with their angle. This article teaches you to spot when the window is misleading.

Quick definition: A rolling average is the average of data over a specific window of time (e.g., the last 30 days, the last 12 months). The window is arbitrary—you can average over any time period you choose. Changing the window changes the average. Headlines exploit this by choosing windows that support their narrative.

Key takeaways​

  • The choice of time window is arbitrary — data over the last month looks different from data over the last year because the window changes what's included
  • Headlines cherry-pick favorable windows — favorable stories use windows where recent data is good; unfavorable stories use windows where recent data is bad
  • Opposing headlines can be true simultaneously — "Stocks rise 20% this quarter" and "Stocks down 30% year-over-year" are both factual if the window changed
  • Longer windows are generally more honest — a one-year average includes more data and is harder to manipulate than a three-month average
  • Cyclical data is most vulnerable to window manipulation — unemployment, inflation, earnings, and other cyclical metrics look different depending on where you start and stop
  • The starting point matters as much as the ending point — if you start measuring from a low point, everything looks like an improvement

Why Windows Matter: The Same Data, Opposite Stories​

Imagine a company's monthly revenue data:

  • January: $10 million
  • February: $12 million
  • March: $11 million
  • April: $13 million
  • May: $14 million

You could report these numbers in multiple ways, all accurate:

"Revenue Surges 40% Over Two Months" (measured from March to May: $11M to $14M)

"Revenue Grows Only 4% Over Four Months" (measured from February to May: $12M to $14M)

"Revenue Flat Year-Over-Year" (if you compare to the same months last year, and they were $14M in May)

All three statements are factually true. But they create completely different impressions. The first suggests rapid acceleration. The second suggests sustained but slow growth. The third suggests stagnation.

The numbers haven't changed. The window changed. A journalist writing a story about company momentum would naturally highlight the 40% over two months. A journalist writing about slowing growth would highlight the 4% over four months. Both are using the same data.

This is window manipulation in action. It's perfectly legal. It's not technically a lie. But it's deceptive.

Here's a real example: Employment data in the United States is reported monthly. Each month, the Bureau of Labor Statistics releases the number of jobs created or lost.

In June 2023, headlines varied:

  • "Unemployment Rises 0.3 Percentage Points in One Month" (measured month-to-month, negative framing)
  • "Unemployment Near Historic Lows, Down 1.2 Points From Pandemic Peak" (measured from 2020 peak, positive framing)
  • "Job Growth Steady, Above Pre-Pandemic Trend" (measured versus historical average, positive framing)

All three were describing the same data point. But the headlines create entirely different impressions depending on the window chosen.

Cherry-Picking the Favorable Window​

Financial journalists don't decide windows randomly. They choose windows that support their angle. If the market is up this quarter, that's the window they lead with. If the market is down this quarter but up year-to-date, they'll use the year-to-date window. If year-to-date is down but the recent month is up, they'll use the recent month.

The process is often unconscious. A journalist genuinely believes the market is recovering, so they naturally reach for the most recent three-month window (where things look better) rather than the year-to-date window (where things look worse). Confirmation bias determines window choice.

But sometimes it's conscious. A financial outlet wants to run a positive story about an asset class, industry, or economy, so they deliberately choose a window that makes things look good.

For example, in early 2023, many outlets ran headlines like "Tech Stocks Rebound Strongly, Best Quarter in Years." This was technically true—tech had an excellent first quarter of 2023. But the three-year or five-year context was very different: tech stocks had collapsed in 2022 and were only recovering from those losses, not reaching new highs. The headline focused on the recent rebound (positive) while ignoring the multi-year context (technology sector valuations were still below 2021 peaks).

The opposite happens when outlets want to run a negative story. In 2022, outlets used multi-year windows to show how bad 2022 was: "Stock Market Down Most in Decade" or "Bonds Have Worst Year in 40 Years." These were true, but they highlighted the window that made things look worst. If you measured from 2008, the market looked great. If you measured from 2021, the market looked terrible. Window choice determined narrative.

Skilled headline writers know this. They choose windows strategically.

Cyclical Data is Most Vulnerable​

Cyclical data is most vulnerable to window manipulation because the choice of starting and ending points dramatically changes the trend.

Unemployment is highly cyclical. It spikes during recessions and falls during booms. You can report unemployment data in a way that makes the economy look great or terrible depending on your window:

"Unemployment Falls Significantly: Down to 3.8% from 4.1% Last Quarter" (recent window, looks good)

"Unemployment Nearly Doubled from 2019 Lows: Up from 2% Before Pandemic" (longer window, looks bad)

Both are true. But the first makes it sound like conditions are improving. The second makes it sound like conditions are worsening. The reality is that the economy is recovering from the pandemic but still hasn't reached the ultra-low unemployment of 2019. Neither headline captures that nuance.

Stock returns are also highly cyclical. A stock market index that's down 20% year-to-date but up 15% in the last month can be reported as either a failure or a recovery depending on the window. A bond market that's had its worst year in 40 years can also be heading into a bull market—context about interest rate cycles matters.

Earnings growth is cyclical too. A company with flat earnings month-over-month but growing earnings year-over-year can be reported as either stagnant or growing depending on the window. A company growing year-over-year but declining quarter-over-quarter can be reported as either a winner or losing momentum.

The safest window for reporting cyclical data is the full cycle—typically 5 to 10 years. Shorter windows will distort the trend by starting or ending at peaks or troughs.

Real-World Examples​

Bitcoin and Cryptocurrency Volatility

Bitcoin has been incredibly volatile. Depending on the window, you can tell completely opposite stories about the same asset:

"Bitcoin Hits New All-Time High: Up 65% This Year" (measured from January 2023 lows)

"Bitcoin Remains 70% Below Peak: Investors Still Down from 2021 Highs" (measured from 2021 peak)

Both were true simultaneously in 2023. Bitcoin had risen substantially year-to-date but was still way below its prior peak. A headline selecting one window (recent recovery) versus another (multi-year decline) would create opposite impressions. Most outlets used whichever window supported whether they were bullish or bearish on crypto.

Real Estate Market Cycles

In 2022–2023, real estate headlines varied wildly depending on window:

"Home Prices Fall 8% as Housing Market Cools" (measured year-over-year, showing recent decline)

"Home Prices Up 30% from Pandemic Lows" (measured from 2020, showing multi-year appreciation)

"Home Prices Down 25% from Peak: Median Home Price Back to 2019 Levels" (measured from 2022 peak)

All three are accurate descriptions of the same market. But they tell stories of decline, recovery, or intermediate status depending on the window chosen.

The S&P 500 in Late 2022 and Early 2023

In late 2022, stock indices were down substantially for the year. Headline: "Stock Market Posts Worst Year in 15 Years."

By early 2023, stocks rallied strongly in January and February. Headline: "Stock Market Surges, Best Start to Year in Decades."

Neither headline was wrong, but they were measured over different windows. The same index that had its worst year in 15 years suddenly had its best start to a year in decades. The change in the index was real. The change in narrative was entirely due to the window shifting.

Unemployment Data During the COVID Recovery

In 2021–2023, unemployment data was reported via multiple windows:

"Unemployment Lowest Since 1969: 3.5%" (recent window, showing extraordinary tightness)

"Unemployment Still Above Pre-Pandemic Levels for Some Demographics" (longer window, showing incomplete recovery)

"Unemployment Fell 10 Percentage Points in Two Years: Fastest Recovery Ever" (measured from peak, showing improvement speed)

All true. All using different windows. Outlets selected windows based on whether they wanted to emphasize strength (recent low) or persistent weakness (pre-pandemic comparison) or recovery speed (peak-to-trough).

How to Evaluate Windows​

When you see a headline with a specific time window, ask yourself:

Is this the natural window, or a cherry-picked window?

Natural windows include:

  • Month-over-month (how did the most recent month compare to the prior month?)
  • Year-over-year (how does this month compare to the same month last year?)
  • Quarter-over-quarter (how does the quarter compare to the prior quarter?)
  • Year-to-date (how does the data look from January 1 through today?)

Cherry-picked windows include:

  • "Since the low" (starting from a recent bottom, making any recovery look huge)
  • "Since the peak" (starting from a recent top, making any decline look huge)
  • "Best week in X years" (choosing a window where one week looks exceptional)
  • "Worst month in X years" (choosing a window where one month looks disastrous)
  • Custom windows designed to show a favorable or unfavorable trend

If a headline uses a cherry-picked window, the journalist is trying to manipulate your perception. That doesn't mean the data is false—it just means the window was chosen to support a narrative.

Is the full picture consistent with this window?

If you see "Market Posts Best Month in 20 Years," ask: Is the market up year-to-date? Up three-year? Or is one exceptional month hiding a bad longer-term trend? A truly strong market will look good over multiple windows (recent month, quarter, year, five-year). A cherry-picked story will look good over one window but not others.

Does this window answer the question I actually care about?

If you're trying to evaluate long-term trends, month-to-month data is noise. If you're trying to understand short-term signals, year-over-year data might be too smoothed. Choose windows that match your investment horizon.

Window Evaluation Decision Tree​

Common Mistakes​

Mistake 1: Assuming a headline's window is neutral. Every headline chooses a window, and that choice is editorial. If you see "Best Quarter in 5 Years," the journalist is choosing a window that makes things look good. Always ask: What would the story look like over a different window?

Mistake 2: Not checking the starting point. A story that "falls 50% from peak" is very different from one that "rises 50% from trough," even though the magnitude is the same. Where the window starts matters as much as where it ends.

Mistake 3: Accepting "year-to-date" without questioning seasonality. Year-to-date is a natural window, but it's not neutral for seasonal data. Employment is typically higher in summer. Retail sales are higher in November–December. A "strong year-to-date" might just be normal seasonality.

Mistake 4: Confusing short-term rebounds with long-term trends. A market that crashes 30% then recovers 10% (still down 20%) can be reported as "surging" (recent window) or "down" (full window). Recent rebounds in cyclical markets are noise; the trend is what happens over the full cycle.

Mistake 5: Ignoring multiple windows for the same metric. If you see an unemployment headline, automatically check year-over-year and three-year comparisons. If unemployment is down month-over-month but up year-over-year, that's more informative than either window alone.

FAQ​

How do I know the "right" window to use?​

It depends on your question. If you're evaluating a long-term investment (buy and hold for 10 years), look at 5–10 year windows. If you're trying to spot a trend change, look at 1–3 year windows. If you're trying to understand monthly volatility, look at month-to-month. Most headlines should include year-over-year (same period last year) because it accounts for seasonality and inflation.

Are longer windows always better?​

Longer windows are more robust to cherry-picking because they include more data. But they can also hide important turning points. A 30-year stock market average hides the 2008 crash and the 2020 rebound. Use multiple windows: recent (last month), medium (last year), and long-term (five years or more).

If a headline uses a favorable window, does that mean it's lying?​

Not lying, but misleading. The data is factually correct. But the window was chosen to support a narrative. An honest article would include multiple windows or explicitly acknowledge that the favorable window is short-term and broader context is different.

How do I compare headlines from different outlets that use different windows?​

It's difficult. One outlet reports "Best Quarter in 10 Years" (recent window) while another reports "Market Still Down from Peak" (all-time window). Both can be true simultaneously. The safest approach is to verify the actual data yourself: What's the absolute return over the last 1, 3, and 5 years? Then you're not dependent on any outlet's window choice.

Can I use rolling averages to smooth out noise?​

Yes, but rolling averages introduce their own biases. A 30-day rolling average will look different from a 90-day rolling average, and headlines can cherry-pick which rolling average to emphasize. The general principle is: longer rolling averages reduce noise but hide short-term turning points. Shorter rolling averages capture turning points but include more noise.

Why do headlines report "since the low" or "since the peak"?​

Because those windows maximize the magnitude of the move. A stock that trades from $50 to $75 (50% gain) sounds less impressive than the same stock reported as "up 150% from its low of $30." The second window is cherry-picked to maximize the percentage, but it's technically accurate if the low really was $30.

Summary​

Financial headlines report data over arbitrary windows of time, and the choice of window dramatically changes the story. A stock that's down 20% year-to-date but up 15% in the last month can be reported as either a failure or a recovery depending on the window. Journalists choose windows strategically—favorable windows when they want positive stories, unfavorable windows when they want negative stories. Cyclical data like unemployment, stock returns, and earnings are most vulnerable because starting and ending points change the trend dramatically. Natural windows (month-over-month, year-over-year, quarter-over-quarter, year-to-date) are harder to manipulate than cherry-picked windows ("best quarter in 5 years," "since the low"). To evaluate a headline, always ask whether the window is natural or cherry-picked, and check how the story would change over different windows. Longer windows are more robust, but using multiple windows (recent, medium-term, long-term) gives you the most complete picture.

Next​

→ Moving averages explained