Skip to main content

How Cherry-Picking Data Shapes Financial News Narratives

A financial analyst publishes a report claiming that technology stocks are undervalued. They compare the price-to-earnings ratio of tech stocks to the historical average and show that current valuations are below the long-term median. Based on this data, they recommend buying tech stocks aggressively.

What they don't mention is that they chose the 30-year historical average as their benchmark. If they had chosen the 10-year average, tech stocks would appear expensive. If they had chosen the 5-year average, they would appear extremely expensive. They selected the time period that made their case, and presented it as objective analysis.

This is cherry-picking data—selecting specific data points, time periods, or metrics that support a desired conclusion while ignoring other data that contradicts it. It's one of the most common and damaging forms of bias in financial news. It's also one of the hardest to detect, because the selected data is often genuinely accurate. The problem isn't that the data is wrong; it's that you're not seeing the full picture.

Cherry-picking is particularly prevalent in financial news because there's always a narrative to support. An article needs a point of view. An analyst needs a recommendation. A headline needs an angle. Selecting data that supports the narrative is easy, natural, and almost always unintentional. The person doing it often isn't even aware they're doing it.

Understanding cherry-picking is essential to reading financial news critically. Without it, you'll absorb a systematically distorted view of asset valuations, strategy effectiveness, and market conditions.

Quick definition: Cherry-picking data in financial news is the practice of selecting specific data points, time periods, metrics, or statistics that support a particular narrative or recommendation while ignoring other data that would point to different conclusions—distorting the apparent truth of the situation.

Key takeaways​

  • Time period selection is the most common form of cherry-picking — choosing when to start and end your analysis radically changes the conclusions
  • Metric selection matters enormously — using price-to-earnings instead of price-to-book, or current valuations versus historical highs, points to different conclusions
  • Survivorship bias and cherry-picking combine — you see successful strategies from periods when they worked, not when they didn't
  • Geographic and sector selection skews analysis — choosing which countries or industries to analyze affects conclusions
  • Scale and comparison points are chosen to support narratives — comparing a stock to the wrong benchmark makes bad stocks look good
  • Cherry-picking often happens unconsciously — analysts genuinely believe their narrative, not realizing they're ignoring contradictory data
  • Detecting cherry-picking requires asking specific critical questions — what other time periods, metrics, or comparisons would tell a different story?

How Time Period Selection Creates Narratives​

The most powerful form of cherry-picking in financial news is time period selection. Different time periods tell completely different stories.

Consider US stock market returns. Here's what you might read in financial news depending on the time period selected:

  • "Stocks have delivered 10.2% annual returns over the past 90 years" (1934-2024)
  • "Stocks have delivered only 3.8% annual returns over the past 15 years" (2009-2024)
  • "Stocks have delivered 14.7% annual returns over the past 5 years" (2019-2024)
  • "Stocks have delivered -5.6% annual returns over the past year" (2023-2024)

All of these statements are true. But which one you highlight depends on the narrative you want to tell. If you want to be bullish, you highlight the 5-year return. If you want to be bearish, you highlight the 1-year return.

The same data can support radically different conclusions depending on which slice you report.

A real example from financial media in 2023: A news outlet ran a story claiming "Tech stocks have underperformed the market." They compared tech stock returns to the broader market return from January 2023 to December 2023.

But if you extended the time period just slightly—say, from January 2020 to December 2023—the opposite was true: tech stocks dramatically outperformed. The outlet had selected a time period where tech was weak and ignored the period where it was strong.

This isn't accidental. When you write an article saying "Tech stocks are weak," you're implicitly claiming the current period is notable. You select a time period where that's true and report it.

The reader sees the article and thinks, "I didn't know tech stocks were weak." But they might be weak only in the selected time period. In a longer period, they might be very strong. The article created a false impression through time period selection.

Metric Selection: How Which Numbers You Use Shapes Conclusions​

Beyond time periods, analysts cherry-pick which metrics to report.

A stock is trading at $50. Different metrics tell different stories:

  • Price-to-earnings ratio: 15x (cheap compared to historical average of 17x)
  • Price-to-book ratio: 2.8x (expensive compared to historical average of 1.5x)
  • Price-to-sales ratio: 1.2x (cheap compared to historical average of 1.4x)
  • Dividend yield: 1.8% (below the historical average of 2.2%, suggesting it's expensive)

An analyst wanting to make a case that the stock is cheap might highlight the price-to-earnings ratio and price-to-sales ratio. An analyst wanting to claim it's expensive would highlight the price-to-book ratio and dividend yield. Both are being truthful about which metrics they report. They're just selecting the metrics that support their conclusion.

A skilled analyst reporting fairly would present all four metrics and explain that valuations are mixed—some suggest the stock is cheap, others suggest it's expensive. But that doesn't make for a compelling recommendation. A bullish or bearish conclusion requires selecting metrics that support the direction.

Here's a real example: In 2021-2022, as growth stocks fell sharply, some analysts argued that they were still expensive. Others argued they were becoming reasonably valued. They were often looking at the same stocks. The difference was metric selection.

Those claiming growth stocks were still expensive were comparing current valuations to long-term historical averages. Those claiming they were becoming cheap were comparing current valuations to the peak valuations of 2021. Both comparisons were valid. The selection of which comparison to highlight created the narrative.

Benchmark Selection: What You Compare Against​

Another form of cherry-picking is selecting which benchmark to compare against.

A mutual fund manager publishes a report showing that their fund returned 8% annually over the past 10 years. This sounds good. Is it?

Well, if the S&P 500 returned 12% annually over the same period, then the fund significantly underperformed. If the S&P 500 returned 5% annually, then the fund significantly outperformed.

The fund's absolute return (8%) is identical either way. But the comparison point you choose—the benchmark—determines whether you interpret it as success or failure.

Fund managers often select benchmarks that make their performance look good. A healthcare-focused fund might compare itself to the healthcare sector's return (where it might have underperformed less badly) rather than to the broader market. A small-cap fund might compare itself to a small-cap index (where it might have beaten) rather than to a large-cap index or the broad market.

The SEC requires funds to disclose how they compare to their benchmark. But if the fund manager chose an advantageous benchmark, the comparison looks good despite potentially underperforming better alternatives.

This affects your investment decisions. If you read that a healthcare fund outperformed its benchmark, you might think it's a good fund. But if you compared it to a broad market index, it might have underperformed. The benchmark selection created the narrative.

Geographic and Sector Selection​

Cherry-picking also extends to which geographic regions and sectors you analyze.

A financial advisor claims that "dividend-paying stocks have delivered superior returns." They might be selecting:

  • A time period where dividend stocks outperformed (say, 2010-2015)
  • Countries where dividend stocks were favored (say, developed markets in Europe)
  • Sectors where dividends correlated with returns (say, utilities and financials)

If they analyzed the same question from 2015-2023, globally, across all sectors, they might find that dividend stocks underperformed. But by selecting the time period, geography, and sectors where dividend stocks performed well, they support their claim.

A real example: Some analysts claim "Value investing is dead." They're selecting a recent time period (2010-2023) where value significantly underperformed growth. But if you extended the analysis back to 1980, value and growth delivered similar long-term returns. The claim that "value is dead" is true only in the selected time period.

Example: How Multiple Forms of Cherry-Picking Combine​

To see how cherry-picking works in practice, consider a real example from financial news.

In 2023, an analyst published a report: "Real Estate Investment Trusts Offer Superior Value." The report included the following data:

  • REIT valuations compared to their 20-year historical average (REITs were below average)
  • REIT dividend yields compared to their 20-year historical average (yields were above average)
  • REIT returns compared to stock returns over the past 5 years (REITs had outperformed)
  • REIT sector selection focused on residential and industrial REITs (best performers), excluding struggling office REITs

What the report didn't include:

  • REIT valuations compared to the current forward earnings growth rate (revealing they were fairly valued, not cheap)
  • Historical periods where REITs traded below average valuations and subsequently underperformed for years
  • The duration of REIT dividend yields (which had increased due to rising interest rates, creating reinvestment risk)
  • REIT returns compared to stocks over the past 20 years (where stocks had significantly outperformed)
  • The struggling office REIT sector

By selecting favorable time periods, metrics, and sector data, the analyst made a case that REITs offered superior value. A different selection of the same data would have suggested caution.

The analyst likely wasn't being deceptive. They probably genuinely believed REITs were attractive and selected the evidence that supported this view, unconsciously ignoring the evidence that contradicted it.

How Cherry-Picking Happens Unconsciously​

Most cherry-picking in financial news isn't deliberate. Instead, it results from how human cognition works.

Once you form an opinion—"Tech stocks are overvalued" or "The market is headed for a crash"—you unconsciously seek information that confirms your view. This is confirmation bias, which you've read about. But confirmation bias often manifests as cherry-picking: you seek out data that supports your view and downplay data that contradicts it.

An analyst might write a report saying "Tech stocks are overvalued based on current valuations." While writing, they collect data supporting this: current price-to-earnings ratios are above the 20-year average, the Shiller cyclically adjusted PE ratio is elevated, growth rates are slowing. They include this data in their report.

But there's also data that suggests tech stocks might be reasonably valued: they've disrupted their industries with lasting advantages, earnings growth is still positive (just slower), the 5-year price-to-earnings average is below the 20-year average. Did the analyst include this? Probably not, not because they're being deceptive, but because they didn't seek it out as actively. They were confirming their already-formed belief that tech is overvalued.

This happens in almost all financial analysis. The analyst starts with a view. They seek data supporting that view. They emphasize the supporting data and downplay contradictory data. The process feels like objective analysis to them—they're just reporting the relevant data—but it's actually cherry-picking.

Real-world examples: Cherry-Picking in Famous Analyses​

Example 1: Housing Market Analysis (2006) Many analysts claimed housing was undervalued in 2006, despite prices having risen dramatically. They selected metrics showing:

  • Housing as a percentage of income was lower than in the 1980s (true, but cherry-picked—it was higher than in the 1990s)
  • Predicted housing demand from demographic trends (true, but ignored that supply was already being built)
  • Historical price appreciation rates (which would have looked obviously excessive with a different time period)

They didn't emphasize:

  • Housing prices compared to historical multiples of rent (revealing excessive valuations)
  • The growth rate in housing supply (which would exceed demand growth)
  • Subprime lending growth and its implications

Example 2: Bitcoin Analysis (2017) During Bitcoin's surge to $20,000, analysts made bullish cases by comparing:

  • Bitcoin's growth rate to early internet adoption (but ignoring that cryptocurrency adoption was much slower)
  • Bitcoin's market cap to gold's market cap (implying Bitcoin could reach gold's value, but ignoring fundamental differences)
  • Bitcoin's use cases (emphasizing its potential while ignoring that it wasn't actually being used for most of these cases)

They didn't emphasize:

  • Bitcoin's price history of multiple crashes from previous peaks
  • The lack of intrinsic cash flows or earnings to justify valuation
  • The regulatory risks and technical limitations

Example 3: Tech Stock Valuations (2020-2021) During the pandemic, tech stocks surged. Analysts used metrics like:

  • Price-to-sales ratios (which grew extremely high)
  • Forward earnings multiples at assumed high growth rates (which didn't materialize)
  • Comparisons to previous tech booms with different characteristics

But ignored:

  • Price-to-current earnings (which was extremely high)
  • Historical precedent for such valuations (which typically fell sharply)
  • The unsustainability of the assumed growth rates

How to Detect Cherry-Picking​

The key to detecting cherry-picking is asking: "What data would I see if different choices had been made?"

When you read financial analysis, ask these questions:

About time periods:

  • Why did the analyst start and end the analysis where they did?
  • What would the conclusion be if you used different time periods?
  • Would a longer history support the same conclusion?
  • Would just the past year or two support the opposite conclusion?

About metrics:

  • What other metrics could measure the same thing?
  • Are there metrics that would point to a different conclusion?
  • Did the analyst mention why they chose this metric over others?
  • Does the metric have known limitations they didn't mention?

About benchmarks:

  • What alternative benchmarks could you use?
  • Would different benchmarks change the conclusion?
  • Is the chosen benchmark representative of alternatives?

About scope:

  • Did the analyst select specific sectors, geographies, or stocks?
  • Would including all sectors/geographies change the conclusion?
  • Are the selected examples representative?

About comparisons:

  • What is being compared to what?
  • Are the compared items actually comparable?
  • Would different comparison points change the conclusion?

Common mistakes: Not Realizing How Many Degrees of Freedom Analysts Have​

A fundamental mistake in reading financial analysis is not realizing how many choices analysts make about which data to present.

An analyst has decisions to make about:

  • What time period to analyze (50 years, 20 years, 10 years, 5 years, 1 year?)
  • What metrics to use (P/E, P/B, P/S, dividend yield, free cash flow yield?)
  • What benchmarks to compare against (broad market, sector, international?)
  • What geographic regions to include
  • What sub-sectors to analyze
  • What comparison points to make (current price vs. historical average, vs. peak, vs. trough?)

With so many degrees of freedom, an analyst can almost always find a selection of data that supports their conclusion. This is true even if the analyst is trying to be objective. The process of writing a report requires making choices, and those choices inevitably influence the conclusion.

If you're seeing a conclusion, and the analysis supports it, it might be because the analysis was careful and objective. But it might also be because the analyst had 50 different ways to analyze the data and selected the analyses that supported their preferred conclusion.

FAQ: Cherry-Picking and Data Interpretation​

Is it possible to present data without cherry-picking?​

Not entirely. Any analysis requires selecting what to include and what to exclude. But you can minimize cherry-picking by: presenting multiple time periods and metrics, acknowledging limitations and alternative interpretations, explaining why certain choices were made, and being transparent about what data you're not including and why.

How do I distinguish between legitimate metric selection and cherry-picking?​

If an analyst explains why they chose their metrics and acknowledges that other metrics might point to different conclusions, they're being thoughtful. If they present one metric without mentioning alternatives, they're likely cherry-picking. The most objective analyses present multiple metrics and explain where they disagree.

Are financial institutions more likely to cherry-pick if they have an incentive in the conclusion?​

Yes, substantially. A bank that wants to encourage you to buy a stock might cherry-pick optimistic data. An analyst who's publicly bullish on a stock and wants to maintain their reputation for being right might cherry-pick supporting data while ignoring contradictory data. Be especially skeptical of analyses where the author has a financial incentive in the conclusion.

How should I respond when I see one-sided analysis?​

Either seek out the opposite perspective from a source without a conflicting incentive, or gather multiple metrics and time periods yourself to get a fuller picture. If one analyst says "Tech stocks are overvalued" and selects data supporting that, find another analyst who evaluated the same stocks and see what they emphasize. The contrast often reveals cherry-picking.

Can historical data be cherry-picked?​

Yes. Choosing which historical period to look at radically affects conclusions. A stock that underperformed for 5 years might have outperformed for the prior 5 years. Both are real data, but different time periods point to different conclusions about how good or bad it is.

How do professional investors protect themselves against cherry-picking?​

They diversify their information sources, look for data that contradicts their views, track their own predictions and update them when wrong, and maintain healthy skepticism of any analysis that's too neat or too certain. They also track which analysts' cherry-picked analyses have been right or wrong over time, downweighting analysts who are frequently wrong despite the analysis appearing logical.

If multiple analyses cherry-pick differently, how do I decide which to believe?​

Look for which conclusions align with the longest time periods and broadest sets of metrics. If one analysis requires a very specific time period and metric selection to support the conclusion, while another conclusion holds across multiple time periods and metrics, the second is probably more reliable.

Summary​

Cherry-picking data is one of the most pervasive biases in financial news. Analysts have countless choices to make about what data to present: which time period to analyze, which metrics to highlight, which benchmarks to use, which sectors to include, and what comparison points to make. By selecting data points that support a predetermined conclusion while omitting data that contradicts it, analysts can make almost any argument seem well-supported, often without realizing they're doing it. The reader, seeing the supporting data and not the omitted data, forms a skewed impression of the situation. Protecting yourself requires asking what data would point to a different conclusion and seeking analyses that present multiple perspectives and metrics rather than relying on single-perspective analysis.

Next​

→ Hindsight bias in news coverage