Tracking Estimate Accuracy Over Time
Tracking Estimate Accuracy Over Time
Some analysts consistently estimate earnings more accurately than others. Over a 5-year period, one analyst might have 58% of estimates within 5% of actual results while another analyst is within 5% only 38% of the time. This variation is not random. Analysts who are more accurate typically have better data sources, stronger research processes, and less sell-side bias. For investors, tracking which analysts are accurate and which are biased is a way to increase the signal-to-noise ratio in analyst forecasts.
Quick definition: Estimate accuracy measures how close an analyst's forecasted earnings per share (or revenue) is to the actual reported results. Accuracy is typically measured as percentage error. A forecast that misses by 5% is more accurate than one that misses by 15%. Over time, an analyst's accuracy record reveals their research quality and potential biases (consistently too bullish or too bearish).
Key takeaways
- Accuracy varies dramatically by analyst: Some analysts achieve 55–60% of estimates within 5% of actual; others achieve 35–40%
- Top-ranked analysts are not always most accurate: Institutional Investor ranked analysts (based on poll of money managers) are often less accurate than lesser-known analysts with better research processes
- Bias is more visible than accuracy: Most sell-side analysts are systematically biased high (consistently too bullish). Tracking this bias is more useful than tracking raw accuracy
- Accuracy persists: Analysts who were accurate in the past two years are likely to be accurate in the next two years; vice versa for inaccurate analysts
- Accuracy metrics drive consensus value: Databases like FactSet and StarMine use accuracy metrics to weight analyst estimates; the most accurate analysts have more influence on consensus
- Sector-specific accuracy matters: An analyst accurate in tech might be inaccurate in healthcare. Accuracy is not transferable across sectors
What Estimate Accuracy Means
Estimate accuracy is measured in several ways:
Absolute Percentage Error (APE)
The most common measure is APE, which calculates the absolute difference between estimated and actual EPS as a percentage of actual EPS.
Formula: APE = |Estimated EPS - Actual EPS| / |Actual EPS| × 100%
Example: Analyst estimates $2.50 EPS. Company reports $2.60. APE = |2.50 - 2.60| / 2.60 = 3.8%
An analyst with an average APE of 5% is accurate. An analyst with an average APE of 15% is less accurate.
Percentage Accuracy (Within X%)
Another metric is "percentage of estimates within X% of actual." This is intuitive: how often is the analyst within 5% of actual? Or within 10%?
Example: Analyst's last 20 quarterly EPS estimates:
- 14 within 5% of actual = 70% accuracy within 5%
- 18 within 10% of actual = 90% accuracy within 10%
This is how research databases report accuracy. A "Top 10% Analyst" typically has 55–65% of estimates within 5% of actual. An average analyst achieves 45–55%. A poor analyst achieves 35–45%.
Bias (Systematic Over/Underestimation)
Bias is whether the analyst is systematically too bullish (overestimates) or too bearish (underestimates). An analyst with 50% APE but biased 10% high is less useful than an analyst with 8% APE and zero bias.
Example: Analyst's last 20 quarterly estimates are on average 8% too high (biased bullish). Every estimate misses high. This bias is more valuable to know than the average magnitude of error.
How to Measure an Analyst's Accuracy
Most investors don't track accuracy manually. Services like FactSet, StarMine, and Morningstar publish accuracy statistics for individual analysts and consensus. However, you can calculate accuracy yourself for any analyst covering a stock you own.
Manual Tracking: Spreadsheet Method
Create a spreadsheet with:
| Quarter | Analyst EPS Est. | Actual EPS | APE | Within 5%? |
|---|---|---|---|---|
| Q1 2024 | $1.50 | $1.48 | 1.4% | Yes |
| Q2 2024 | $1.65 | $1.58 | 4.4% | Yes |
| Q3 2024 | $1.72 | $1.61 | 6.8% | No |
| Q4 2024 | $1.85 | $1.92 | 3.6% | Yes |
After 8 quarters of data, calculate:
- Average APE: 4.1%
- Percentage within 5%: 75%
- Average bias: Analyst was 1.2% high on average (bullish bias)
This simple tracking reveals the analyst's accuracy pattern. If you see the analyst is consistently 2–3% too high, you can mentally adjust future estimates downward.
Digital Tools for Tracking
Several services publish analyst accuracy statistics:
- FactSet: Provides accuracy rankings for individual analysts across metrics (APE, median absolute percentage error, bias)
- StarMine: Owned by Thomson Reuters; identifies "Star" analysts based on estimate accuracy and stock-picking ability
- Morningstar: Publishes accuracy data for analysts covering specific stocks
- TradingView: Shows analyst accuracy for major stocks
- Refinitiv Data: Comprehensive database of analyst estimates and actual results
These services weigh estimates by analyst accuracy when calculating consensus. If Analyst A is accurate 60% within 5% and Analyst B is accurate 40% within 5%, Analyst A's estimate is weighted more heavily in consensus.
The Mermaid: Accuracy Tracking Over Time
Who Are the Most Accurate Analysts?
Research on analyst accuracy reveals consistent patterns. Most sell-side analysts are systematic biased high (too bullish). The most accurate analysts often come from:
1. Independent Research Firms
Analysts at firms like Morningstar, Altimeter, and independent boutiques often have higher accuracy than bank analysts. Why? Because they have no investment banking relationships with the companies they cover, so they have no pressure to be bullish.
Example: A Morningstar analyst covering Intel has no banking relationship with Intel. If Intel's valuation is expensive, the analyst can say so without worrying about losing a banking client. A bank analyst at Morgan Stanley might be cautious about being too negative on Intel because Intel might use Morgan Stanley for M&A advice.
2. Specialist Analysts in Niche Sectors
An analyst covering 8 biotech companies knows the industry deeply. An analyst covering 40 healthcare companies has more breadth but less depth. The specialist is often more accurate because they understand nuances of the sector.
3. Analysts at Smaller Investment Banks
Analysts at smaller banks (Piper Sandler, Oppenheimer, Evercore) sometimes have better accuracy than analysts at mega-banks (Goldman Sachs, Morgan Stanley, JP Morgan). Why? Smaller banks have fewer banking clients, so analysts have less pressure to be bullish on every company.
4. Analysts Who Specialize in Earnings Forecasting
Some analysts are generalists (covering valuation, strategy, capital allocation). Others specialize in building detailed financial models. Model-specialists often have better accuracy because they spend more time on financial forecasting.
5. "Bottom-Up" vs. "Top-Down" Analysts
Bottom-up analysts build models from customer/unit economics and work up to company totals. Top-down analysts start with macro assumptions and work down to individual stocks. Bottom-up analysts tend to be more accurate because they're building on primary research, not macro assumptions.
Common Patterns in Analyst Bias
Pattern 1: Systematically Bullish (Most Common)
Most sell-side analysts are biased high by 2–5% on average. They estimate higher revenue growth, higher margins, and faster multiple expansion than actually occurs.
Why? Several theories:
- Optimism bias (humans naturally overestimate positives)
- Incentive structure (bullish analysts are popular; bearish analysts are fired)
- Access bias (management guides analysts high to create buying interest)
- Confirmation bias (analysts believe in the thesis and ignore warning signs)
Pattern 2: Systematically Bearish (Rare)
A small minority of analysts are biased low (too bearish). These analysts might be sector bears (always negative on a particular industry) or contrarians (always skeptical of consensus).
Pattern 3: Accurate in Some Quarters, Biased in Others
An analyst might be accurate during calm markets but biased during turning points. For example:
- Accurate when earnings are stable
- Bullish bias during growth acceleration phases
- Bearish bias during slowdown phases
This pattern suggests the analyst is reactive (slow to incorporate new information) rather than anticipatory.
Pattern 4: Accurate on Large Moves, Biased on Small Moves
An analyst might be accurate when a company beats by 10% but inaccurate when a company beats by 1%. This suggests the analyst captures major trends but misses small details.
Real-World Example: Tracking Tesla Analyst Accuracy
Let's track accuracy for three hypothetical Tesla analysts over 8 quarters:
Analyst A: Morgan Stanley (Bank Analyst)
| Q | Estimate | Actual | APE | Bias |
|---|---|---|---|---|
| Q1'24 | $1.10 | $0.98 | 12.2% | +0.12 |
| Q2'24 | $1.05 | $0.94 | 11.7% | +0.11 |
| Q3'24 | $1.08 | $0.96 | 12.5% | +0.12 |
| Q4'24 | $1.15 | $1.02 | 12.7% | +0.13 |
Metrics:
- Average APE: 12.3%
- Within 5%: 0/4 = 0%
- Average bias: +11.5% (consistently too bullish)
Assessment: Analyst A is consistently bullish. You can expect this analyst to overestimate Tesla earnings by 10–12% on average. To use this analyst's estimates, subtract 12% and you have a more accurate forecast.
Analyst B: Morningstar (Independent)
| Q | Estimate | Actual | APE | Bias |
|---|---|---|---|---|
| Q1'24 | $0.99 | $0.98 | 1.0% | +0.01 |
| Q2'24 | $0.97 | $0.94 | 3.2% | +0.03 |
| Q3'24 | $0.96 | $0.96 | 0.0% | 0.00 |
| Q4'24 | $1.04 | $1.02 | 2.0% | +0.02 |
Metrics:
- Average APE: 1.6%
- Within 5%: 4/4 = 100%
- Average bias: +1.5% (slightly bullish but minor)
Assessment: Analyst B is highly accurate. This analyst's estimates are reliable. You can use this analyst as a primary source for Tesla earnings forecasts.
Analyst C: Altimeter (Bearish House)
| Q | Estimate | Actual | APE | Bias |
|---|---|---|---|---|
| Q1'24 | $0.88 | $0.98 | 10.2% | -0.10 |
| Q2'24 | $0.85 | $0.94 | 9.6% | -0.09 |
| Q3'24 | $0.92 | $0.96 | 4.2% | -0.04 |
| Q4'24 | $0.98 | $1.02 | 3.9% | -0.04 |
Metrics:
- Average APE: 7.0%
- Within 5%: 2/4 = 50%
- Average bias: -6.7% (bearish, underestimating)
Assessment: Analyst C is bearish but improving. Early quarters were inaccurate; later quarters improved. This suggests the analyst was wrong about Tesla's trajectory but eventually adjusted. You can use recent estimates but be cautious about earlier quarters.
Which Analyst to Trust?
Based on the tracking above:
- Best choice: Morningstar (1.6% APE, minimal bias, 100% within 5%)
- Second choice: Altimeter (improving accuracy, though biased low)
- Least reliable: Morgan Stanley (12.3% APE, consistent bullish bias)
If you're trying to forecast Tesla earnings, use Morningstar's estimate as a base and add 2–3% to account for typical bullish bias. Ignore Morgan Stanley's estimate unless you mentally discount it by 12%.
How to Use Accuracy Tracking for Investment Decisions
Insight 1: Use Accurate Analysts' Revisions as Signals
If Analyst B (the accurate one) just revised estimates up 5%, this is a strong signal that fundamentals are improving. If Analyst A (the biased one) revised up 5%, it might just be bullish bias continuing.
Insight 2: Trust Consensus from Accurate Analysts More
If five analysts cover a stock and three are accurate and two are biased, the consensus should weight the three accurate analysts more heavily. Databases like FactSet do this automatically, but you can do it manually by excluding biased analysts from your calculation.
Insight 3: Use Bias to Estimate Actual Earnings
If an analyst is consistently biased 5% high, and they estimate next quarter earnings of $2.00, adjust it to $1.90 for your own analysis.
Insight 4: Monitor for Bias Changes
If an analyst who was biased high for five years suddenly becomes accurate, something changed. Maybe they got a new manager, or they switched industries. This can increase your confidence in their estimates.
Insight 5: Use Accuracy to Build Confidence Ranges
Instead of trusting a single analyst estimate, build a range based on accuracy:
- If the accurate analyst estimates $2.00 with a typical APE of 2%, expect results within $1.96–$2.04
- If the less-accurate analyst estimates $2.10 with a typical APE of 10%, expect results within $1.89–$2.31
The more conservative range from the less-accurate analyst is appropriate.
Common Mistakes: Misinterpreting Accuracy Data
Mistake 1: Assuming high accuracy in one quarter predicts high accuracy next quarter
Analyst accuracy is mean-reverting. An analyst accurate in Q1 might be inaccurate in Q2. Look at 8–12 quarters of data before drawing conclusions.
Mistake 2: Confusing accuracy with utility
An analyst might be inaccurate but systematically biased in a predictable direction. This bias is useful because you can adjust for it. An analyst who is accurate 50% of the time with random error is less useful than an analyst who is inaccurate 30% of the time with consistent bias.
Mistake 3: Penalizing analysts for macro surprises
If an analyst estimated 10% revenue growth but a recession caused 5% growth, was the analyst inaccurate or did macro shift? True accuracy metrics should control for macro shifts, but simple APE calculations don't.
Mistake 4: Using accuracy for the wrong metric
An analyst might be highly accurate on EPS but inaccurate on revenue. If you're trying to forecast revenue, use their revenue accuracy, not their EPS accuracy. Different analysts specialize in different metrics.
Mistake 5: Assuming top-ranked analysts are most accurate
Institutional Investor ranked analysts (based on investor surveys) are popular but not always most accurate. Some top-ranked analysts are overrated due to their personality or access, not their accuracy.
FAQ: Estimate Accuracy
How many quarters of data do I need to assess an analyst's accuracy?
At least 8–12 quarters. With only 4 quarters, one lucky or unlucky quarter dominates the statistics. With 12+ quarters, patterns become clear.
Should I track absolute error or percentage error?
Percentage error (APE) is better because it normalizes for stock price. A $0.10 miss on a $2.00 stock is worse than a $0.10 miss on a $20.00 stock.
Is accuracy more important than ranking?
For your own investment decisions, yes. An accurate analyst not ranked by Institutional Investor is more useful than an inaccurate analyst ranked top 10. Ranking reflects popularity with money managers; accuracy reflects research quality.
How do I find analyst accuracy data?
FactSet and StarMine publish detailed accuracy statistics. Morningstar publishes accuracy for analysts covering specific stocks. TradingView shows simplified accuracy metrics.
Can I beat the consensus by using accuracy weighting?
Possibly. If you identify analysts who are systematically more accurate on a specific metric (e.g., revenue growth) or in a specific sector, you can create a weighted estimate that beats consensus. However, accuracy advantages tend to diminish over time as other investors copy successful strategies.
Do analysts improve their accuracy over time?
Some do, some don't. Analysts who receive feedback (from management or the market) often improve. Analysts insulated from negative feedback often don't improve. Early-career analysts sometimes improve rapidly as they build experience.
Related Concepts
- Where to Find Estimates
- What is the Consensus
- Top-Ranked Analysts
- Analyst Revisions and Momentum
- The Consensus Over Time
Summary
Estimate accuracy varies dramatically across analysts. Some analysts achieve accuracy within 5% more than 60% of the time; others achieve it less than 40% of the time. By tracking which analysts are accurate and which are biased, investors can improve the signal-to-noise ratio in professional forecasts.
The most important insight is that bias is often more useful to know than accuracy. An analyst who is inaccurate but predictably bullish (biased high by 8% on average) is more useful than an analyst who is randomly inaccurate. You can adjust the biased analyst's estimates mentally and increase reliability.
Tracking estimate accuracy requires discipline—maintaining a simple spreadsheet of estimates versus actual results. But the effort pays off. After 12–16 quarters of tracking, clear patterns emerge. You learn which analysts to trust, which to adjust, and which to ignore. This insider knowledge gives you an advantage in forecasting earnings and positioning your portfolio.
The databases that measure analyst accuracy (FactSet, StarMine) weight analysts' estimates according to their historical accuracy when calculating consensus. By understanding how they weight estimates, you can reverse-engineer which analysts are considered most reliable and give their views more credence in your own analysis.
Next
Independent vs. Bank Research