Skip to main content

Why Comparing Raw Data Without Adjustment Creates False Trends

A headline announces: "Retail Sales Plunge 8% in January." It sounds like a consumer spending crisis. The economy is weakening. Stocks sell off.

But here's the context: retail sales always fall sharply in January. Christmas happens in December, and in January, consumer spending drops after the holiday surge. This isn't new, surprising, or a sign of economic weakness. It's a completely predictable seasonal pattern that occurs every single year.

The 8% decline in January compares to December sales (inflated by holiday shopping). The real question is: How do January sales this year compare to January sales last year? If they're up year-over-year, the economy is strengthening. If they're down year-over-year, it's weakening. The month-to-month comparison is meaningless noise.

This is the seasonal adjustment problem. Financial data contains predictable seasonal patterns that repeat annually. Ignoring these patterns and comparing raw numbers creates false signals of strength or weakness. Journalists exploit this by reporting "shocking" declines that are entirely normal, or reporting "strength" when seasonality inflates the numbers.

Quick definition: Seasonal adjustment is the statistical process of removing predictable seasonal patterns from data before comparing year-to-year or making conclusions about trends. A seasonally adjusted unemployment rate has the holiday-hiring effect removed. Seasonally adjusted retail sales have the December surge removed. Raw, unadjusted data looks more volatile and misleading.

Key takeaways​

  • Many economic metrics have predictable seasonal patterns — unemployment dips in summer, retail sales spike in December, agriculture output varies by quarter
  • Comparing month-to-month raw data is almost always misleading — the comparison includes seasonal noise that hides the real trend
  • Year-over-year comparisons control for seasonality naturally — comparing January to January, December to December removes the seasonal distortion
  • Seasonal adjustment is standard in official statistics — the U.S. Bureau of Labor Statistics, Census Bureau, and Federal Reserve publish both raw and seasonally adjusted data
  • Headlines often use unadjusted data to exaggerate moves — a "plunge" in seasonally predictable data sounds worse than a modest decline in seasonally adjusted data
  • Different adjustment methods produce different results — official seasonality adjustments can vary, creating uncertainty about the "true" underlying trend

The Seasonal Cycle: Why Every Month is Different​

Economic activity is not random throughout the year. Humans follow predictable patterns.

Retail sales spike in November and December due to holiday shopping. They plunge in January when holiday spending ends. This pattern repeats identically every year.

Construction activity varies by season. Northern hemisphere construction is higher in spring and summer, lower in fall and winter due to weather. This pattern repeats annually.

Agricultural output varies by growing season. Harvests concentrate in fall. Planting concentrates in spring. This pattern repeats annually.

Hiring and employment are affected by back-to-school hiring in August, summer hiring booms, and holiday hiring in November–December. Layoffs often happen in January after the holidays. This pattern is fairly consistent.

Shipping and logistics show patterns around holidays and major shopping periods. August is the busiest month for many retailers preparing for back-to-school and fall holidays.

Tax receipts are higher in April (tax day) and January (holiday bonus payouts, New Year spending). Lower in other months.

These patterns are predictable, recurring, and well-documented. They're not surprises. But when journalists report raw, unadjusted data, they treat monthly changes as if they're unexpected signals of economic strength or weakness.

The Seasonal Adjustment Process​

To remove seasonal noise, statisticians use a process called seasonal adjustment. Here's how it works in simplified form:

  1. Identify the historical pattern. Calculate the average value for January across the last 10+ years. Same for February, March, etc. This shows you "typically, January is X% below the annual average."

  2. Calculate the adjustment factor. If January sales are typically 12% below the annual average, the adjustment factor for January is 1.12 (multiply unadjusted January data by this factor to normalize it).

  3. Apply to all Januaries. Multiply every January's reported sales by 1.12 to remove the expected January seasonal decline.

  4. Compare adjusted numbers. Now you can compare January (adjusted upward) to December (adjusted downward) and to other months, all on a normalized, seasonally adjusted basis.

The result: seasonally adjusted data removes the predictable patterns and reveals the true underlying trend.

The U.S. government publishes both raw and seasonally adjusted data for most major economic indicators. The unemployment rate, for instance, is reported both ways. Usually, the seasonally adjusted version is what economists focus on, because it's more informative.

But headlines often report raw numbers because they're more volatile and dramatic.

The Trap: Headlines Use Raw Data to Shock​

A journalist sees that retail sales fell 3% from last month. Using raw, unadjusted data, this looks like a decline. Using seasonally adjusted data, maybe retail sales rose 1%. Very different stories.

Which does the journalist report? Often the raw, unadjusted figure, because it sounds more dramatic: "Retail Sales Fall, Consumer Spending Weakens."

If the journalist had reported the seasonally adjusted figure ("Retail Sales Rise 1%, Consumer Spending Steady"), it would be more accurate but less engaging. The raw figure generates more clicks and social media shares.

Here are specific examples of how raw versus seasonally adjusted data tell opposite stories:

December–January retail sales: December has massive retail sales due to Christmas. January always declines steeply. If a journalist compares December to January and reports a month-over-month "plunge," that's predictable seasonality, not a signal of weakness.

Summer unemployment: Many young people enter the workforce in summer looking for jobs. Unemployment often ticks up in May, June, and July. If a journalist reports "Unemployment Rose This Month," it sounds concerning. But it might just be summer entry into the labor force. Seasonally adjusted unemployment might be flat or declining.

Q1 economic growth: Many economists see a "soft patch" in Q1 GDP growth because businesses are rebuilding inventory after the holiday season. Q1 growth is historically slower. If a headline says "Economic Growth Slows in Q1," it might be predictable seasonality, not actual weakness.

A journalist reporting on raw data creates false signals of economic strength and weakness. This misleads investors and the public.

Real-World Examples​

The Seasonal January Employment Collapse

Every January, the Bureau of Labor Statistics releases employment data that shows raw job losses. Seasonally adjusted data usually shows job gains. This is because many workers hired for the December holiday season are laid off in January—a predictable, recurring pattern.

In January 2023, raw employment data showed a loss of about 260,000 jobs. Seasonally adjusted data showed a gain of 517,000 jobs.

Imagine the headline: "Jobs Fall 260,000 in January: Economy Weakening." That's technically accurate for raw data. But investors reacting to that headline would be reacting to predictable seasonality, not real economic change.

If the journalist had used seasonally adjusted data, the headline would be: "Jobs Rise 517,000: Strong Labor Market Continues." Much more accurate. But the journalist chose the raw number, creating false alarm.

Retail Sales Every December and January

December retail sales are the highest of the year due to holiday shopping. They're always the highest. January retail sales are the lowest due to the post-holiday slump. They're always the lowest.

Yet every January, journalists report "Retail Sales Collapse" or "Consumer Spending Plunges." And every December, they report "Retail Sales Soar" or "Holiday Shopping Exceeds Expectations."

These headlines are comparing raw, seasonally volatile data. Using seasonally adjusted figures, the story would often be: December was moderately strong (not as strong as the raw number suggests), and January was modest (not as weak as the raw number suggests).

Construction Starts and Weather

Construction starts vary dramatically by season due to weather. Winter in northern climates has fewer construction starts. Summer has peaks. This is expected and recurring.

Yet headlines report construction data month-to-month without adjustment: "Construction Starts Fall in February" or "Building Surges in June." These are often just seasonality. Seasonally adjusted construction data would show whether the underlying trend is up or down, stripped of the weather effect.

Agricultural Prices and Harvest Seasons

Commodity prices vary by growing season. Corn prices usually fall at harvest (more supply) and rise in non-harvest months (less supply). This is completely predictable.

A headline might say: "Corn Prices Collapse in October" (when the harvest occurs) or "Corn Prices Rise in May" (when supply is scarce). Both statements are predictable seasonal patterns. The real question is: How do prices this harvest compare to last harvest?

Seasonal Data Evaluation Flowchart​

Why Seasonal Adjustment Matters for Investing​

If you're trying to understand economic trends, seasonality is critical.

A stock market that declines in January due to "tax-loss harvesting" (investors selling losers at year-end to offset gains) might recover in February naturally, without any change in the underlying business environment. An investor who sells in panic after a January decline would miss the February recovery.

A company that reports lower Q1 revenue might be experiencing seasonal weakness, not fundamental deterioration. The company might return to normal in Q2. An investor who sells after Q1 weakness without understanding seasonality might sell at the worst time.

Employment data released monthly can be misinterpreted without seasonal adjustment. If you're trying to assess whether the labor market is strengthening or weakening, seasonally adjusted unemployment is far more informative than raw unemployment.

This directly affects investment decisions. Investors who react to headline-grabbing raw data without understanding seasonality are likely to make trades based on noise.

The Adjustment Method Problem​

Here's an additional complication: there are multiple ways to seasonally adjust data, and different methods produce different results.

The official U.S. seasonal adjustments are calculated using the X-13-ARIMA-SEATS algorithm, maintained by the U.S. Census Bureau. This method:

  1. Identifies the seasonal pattern
  2. Extrapolates the pattern forward into the future
  3. Adjusts recent data based on the expected future pattern
  4. Revises past adjustments when new data arrives

Because the algorithm uses forward-looking assumptions, the seasonal adjustments for recent months can be revised when new data arrives. This creates uncertainty—the "true" seasonally adjusted figure might change.

Journalists sometimes report preliminary seasonal adjustments without mentioning that revisions are likely. This creates a false impression of precision.

Alternative adjustment methods (X-11, X-11-ARIMA) produce slightly different results. Private analysts using different methods than the government might report different seasonally adjusted figures.

The point: seasonal adjustment is not a mechanical, objective process. It involves assumptions and methodology choices. But it's far better than reporting raw data and pretending seasonality doesn't exist.

How to Evaluate Seasonal Data​

When you see economic data reported, ask: Is this seasonally adjusted?

If the headline doesn't specify, assume it's raw or unclear. Contact the news outlet and ask. Most official government statistics (BLS, Census Bureau, Federal Reserve) are released both ways.

If you see a big monthly move, immediately ask: Is this seasonality?

Does the moving metric have a known seasonal pattern? (Retail sales: December spike, January collapse. Employment: summer hiring, January layoffs. Construction: winter slowdown. Agriculture: harvest seasonality.)

If yes, don't panic or celebrate the monthly move. Instead, compare year-over-year.

Year-over-year comparison is your friend. If you compare January 2024 to January 2023, you've automatically controlled for seasonality. The seasonal patterns were the same in both years (they're supposed to be), so the difference is real underlying change.

If you compare January 2024 to December 2023, you're confusing seasonality with real change.

Check official sources for seasonal adjustment data. The U.S. Census Bureau publishes detailed seasonally adjusted retail sales. The Bureau of Labor Statistics publishes seasonally adjusted unemployment. The Federal Reserve publishes seasonally adjusted industrial production. These are the authoritative numbers.

Common Mistakes​

Mistake 1: Assuming month-to-month changes are meaningful. Most month-to-month changes in economic data include large seasonal components. Treat these with extreme skepticism unless the news outlet explicitly states the data is seasonally adjusted.

Mistake 2: Comparing December to January without adjustment. December and January are the most seasonally volatile months due to holidays. Any headline comparing them directly is almost certainly misleading.

Mistake 3: Panicking at expected seasonal patterns. Summer unemployment rises due to young people entering the labor force—this is expected every year. Retail sales fall in January—expected every year. Reacting to these as if they're surprises is reactionary.

Mistake 4: Forgetting that seasonal adjustment involves assumptions. Seasonal adjustment is an estimate based on historical patterns. If the pattern changes (due to structural changes in the economy, pandemic disruptions, etc.), the adjustment might be wrong. The data can be "seasonally adjusted" and still misleading if the adjustment method is outdated.

Mistake 5: Not specifying whether you're talking about raw or adjusted data in your own analysis. If you cite "unemployment rose to 4%," specify whether that's raw or seasonally adjusted. The numbers look identical but mean very different things.

FAQ​

Why don't journalists just always use seasonally adjusted data?​

Sometimes they do. But raw data is more volatile and dramatic, which generates engagement. Raw data also requires less explanation (you don't need to say "seasonally adjusted"—the headline is simpler). Lazy reporting defaults to raw data.

If I compare year-over-year, do I need seasonally adjusted data?​

No. Year-over-year comparison naturally controls for seasonality if the seasonal pattern is stable. January 2024 vs. January 2023 removes the seasonal December-to-January decline because both years had that same decline. But year-over-year comparison doesn't work if you're trying to measure short-term trends (month-to-month or quarter-to-quarter). For those, you need seasonal adjustment.

Can seasonal patterns change over time?​

Yes. Seasonal patterns are derived from historical averages, typically 10+ years. If economic behavior changes structurally (e.g., more online shopping shifts the retail seasonality pattern), the historical adjustment factors become outdated. During the COVID pandemic, seasonal patterns shifted because behavior changed. This caused seasonal adjustments to be less reliable.

Why is Q1 economic growth usually slower than other quarters?​

Q1 includes January, which is historically weaker as consumers recover from holiday spending and businesses rebuild inventory after December. This is structural seasonality—businesses and consumers both pull back in Q1 every year. That makes Q1 growth numbers less impressive and more difficult to interpret.

How do I know if a seasonal pattern is "real" or a fluke?​

Look for consistency. If January unemployment has been rising, on average, for 10+ years, that's a real pattern. If January was up a few times and down a few times randomly, that's not a stable seasonal pattern. The Bureau of Labor Statistics uses statistical tests to identify seasonality before adjusting for it.

Do seasonal adjustments work differently for different countries?​

Yes. Countries with different climates and economic structures have different seasonal patterns. Northern hemisphere countries have winter construction slowdowns. Southern hemisphere countries have them in their winter (northern summer). Agricultural countries have different seasonal patterns than service economies. So official seasonal adjustments vary by country.

Summary​

Economic and financial data contain predictable seasonal patterns that repeat annually. Retail sales spike in December and collapse in January. Employment rises in summer and often falls in January. Construction varies by season due to weather. Comparing raw, unadjusted data month-to-month conflates seasonal patterns with real economic change, creating false signals. Journalists often report raw data because it's more volatile and dramatic, generating engagement. To evaluate seasonal data, always ask whether it's seasonally adjusted. If not, use year-over-year comparisons instead, which naturally control for seasonality by comparing January to January, December to December. Official government statistics like unemployment and retail sales are published both ways: raw and seasonally adjusted. The seasonally adjusted version is far more informative for understanding underlying trends. Seasonal adjustment is an imperfect process involving assumptions and methodology choices, but it beats ignoring seasonality altogether. By understanding seasonal patterns and demanding seasonally adjusted data (or using year-over-year comparisons), you can avoid making investment decisions based on predictable noise.

Next​

→ Real vs nominal headlines