Portfolio and risk · Measuring return and risk
Tracking error and information ratio: a fund against its benchmark
How to calculate active return, tracking error and the information ratio, with seven funds measured against SPY and what the numbers say about patience.
On this page
Tracking error is how much a portfolio's return wanders around its benchmark's return from period to period. The information ratio divides the portfolio's average lead over the benchmark by that tracking error. Together they answer two questions about any fund that is not simply an index tracker: how different is it from the index, and has being different paid?
These are the Sharpe ratio's logic applied to a different starting point. The Sharpe ratio treats cash as the alternative and total volatility as the risk. Here the alternative is an index fund, and the risk is the chance of trailing it. For an investor who could have held the index for almost nothing, that is the relevant comparison. This lesson calculates both measures for seven funds against SPY and shows how long it takes for an information ratio to mean anything.
The idea
Take a fund and its benchmark and subtract one from the other every month. What is left is a series of small numbers, some positive and some negative: the months the fund was ahead and the months it was behind. That series is the active return.
Two things can be said about it. Its average says whether the fund was ahead on balance. Its spread says how far apart the fund and benchmark typically were. A fund that copies the index has active returns that are all nearly zero. A fund that holds different stocks has active returns that swing widely, and may spend years on one side.
The figure plots four funds as a ratio to SPY: each fund's growth of $1 divided by SPY's. A flat line at 1 means no difference. A line that rises means the fund is pulling ahead, and one that falls means it is dropping behind.
IVV, another S&P 500 fund, is the flat line: no meaningful active return. The equal-weight fund RSP pulled ahead of SPY for a decade and spent the following decade giving it all back. Tracking error measures how much the lines wiggle. The information ratio asks whether, after all the wiggling, they ended above 1.
How it works
Tracking error is a standard deviation like any other, calculated on the differences instead of the returns. It is quoted in percent a year. The information ratio has no unit.
A worked example on RSP
RSP holds the same 500 companies as SPY in equal amounts, about 0.2% each, where SPY weights them by market value. The data: 262 monthly total returns of each, January 2005 to October 2026.
- Subtract SPY's return from RSP's each month. The 262 differences average −0.003% a month.
- Their standard deviation is 1.57% a month.
- Tracking error: 1.57% × √12 = 5.45% a year.
- Average active return: −0.003% × 12 = −0.04% a year.
- Information ratio: −0.04 ÷ 5.45 = −0.01.
So RSP was, on average, neither ahead nor behind, and in any given year it was typically about five and a half points away from SPY in one direction or the other. Its best year relative to SPY was 2009, when it finished 18.3 points ahead. Its worst was 2023, when it finished 12.5 points behind. It beat SPY in 10 of the 21 full calendar years.
One detail: the compound returns differ by more than the average active return suggests. RSP's CAGR was 0.4 points a year below SPY's, because RSP was also slightly more volatile and volatility lowers compound growth. The information ratio uses the arithmetic average.
Seven funds against SPY
| Fund | From | Active return a year | Tracking error | Information ratio | Years ahead of SPY |
|---|---|---|---|---|---|
| IVV (S&P 500) | Jan. 2005 | +0.03% | 0.38% | 0.07 | 17 of 21 |
| RSP (S&P 500, equal weight) | Jan. 2005 | −0.04% | 5.45% | −0.01 | 10 of 21 |
| VTV (large value) | Jan. 2005 | −1.11% | 5.36% | −0.21 | 6 of 21 |
| QQQ (Nasdaq-100) | Jan. 2005 | +4.24% | 8.04% | 0.53 | 15 of 21 |
| IWM (small companies) | Jan. 2005 | −1.32% | 9.62% | −0.14 | 9 of 21 |
| USMV (low volatility) | Dec. 2011 | −3.08% | 7.38% | −0.42 | 4 of 14 |
| ARKK (active growth) | Dec. 2014 | +5.00% | 27.51% | 0.18 | 7 of 11 |
All returns include dividends and are after each fund's own expenses; the holder's taxes and trading costs are not included.
The tracking errors sort the funds by how unlike SPY they are. The S&P 500 fund is at 0.38%, and even that small figure is mostly dividends being recorded in different months by the two funds. The value and equal-weight funds, which hold the same large companies in different proportions, are near 5.5%. The Nasdaq-100 and small-company funds, which hold different companies, are at 8% to 10%. ARKK, a concentrated portfolio of a few dozen growth stocks, is at 27.5%, nearly twice SPY's own total volatility.
ARKK shows the gap between the two averages at its widest. Its arithmetic active return is +5.0% a year, yet its compound return was 0.3 points a year below SPY's. Extreme swings, a year 134.5 points ahead in 2020 and a year 52.1 points behind in 2021, average out favourably and compound badly.
Using it
Tracking error as a budget for surprise
Because tracking error is a standard deviation, the usual rule of thumb applies: about two years in three, the gap between fund and benchmark should fall within one tracking error, and about nineteen years in twenty within two.
| Tracking error | An ordinary year (two in three) | An unusual but expected year (one in twenty) |
|---|---|---|
| 0.4% | Within 0.4 points of the index | Almost 1 point away |
| 5.5% | Within 5.5 points | 11 points away |
| 8% | Within 8 points | 16 points away |
| 27.5% | Within 27.5 points | 55 points away |
This is the most practical use of the number. An investor who buys a value fund with a tracking error of 5.4% and is alarmed when it trails the index by 9 points in a year has seen nothing unusual. VTV's worst year against SPY was 16.9 points behind (2023) and its best 16.1 points ahead (2022), consecutive years. The time to decide whether that is tolerable is before buying.
Like other volatility estimates, the rule understates the extremes. ARKK's 134.5-point lead in 2020 was far outside two tracking errors.
Is the information ratio different from zero?
An information ratio estimated from a few years is mostly noise. A quick test multiplies it by the square root of the number of years, which gives approximately the t-statistic of the average active return. Values beyond about 2 are unlikely to be chance.
| Fund | Information ratio | Years | IR × √years |
|---|---|---|---|
| QQQ | 0.53 | 21.8 | 2.5 |
| ARKK | 0.18 | 11.9 | 0.6 |
| RSP | −0.01 | 21.8 | 0.0 |
| VTV | −0.21 | 21.8 | −1.0 |
| USMV | −0.42 | 14.9 | −1.6 |
Only QQQ clears the bar, and its lead came from one identifiable source: it holds much more technology than SPY, during two decades in which technology led. That is a persistent exposure that happened to pay, not evidence of selection skill; see Factor models: explaining a portfolio’s returns with several betas.
The same arithmetic run forwards shows how demanding the test is. A manager with a true information ratio of 0.5, which would be considered excellent, needs 16 years of results to reach a t-statistic of 2. With a ratio of 0.25 it takes 64 years.
Tracking difference for index funds
For an index fund, two numbers matter and they are easy to confuse.
- Tracking difference is the average gap: how much the fund trails its index per year. It should be close to the fund's expense ratio.
- Tracking error is the variability of the gap. For a large index fund it should be a small fraction of a percent.
A fund can have a tiny tracking error and a steady tracking difference of −0.5% a year, which means it reliably loses half a point to costs. For an index fund, the difference is the cost and the error is the quality of the replication.
Where it fails or misleads
- Low tracking error is not low risk. IVV has almost no tracking error against SPY and fell just as far in 2008. Tracking error measures risk relative to the benchmark, and the benchmark's own risk is untouched.
- High tracking error is not skill. It only says the fund is different. A concentrated portfolio chosen at random would also have a high one.
- The benchmark must fit. Measuring a small-company fund against SPY gives a tracking error of 9.6%, most of which is the difference between small and large companies. Against a small-company index the figure would be a fraction of that. A fund can look "active" simply by being compared with the wrong index; Alpha: the return that market exposure does not explain has the same weakness.
- It is symmetric. Tracking error counts beating the benchmark as much of a deviation as trailing it.
- It is unstable. Tracking errors rise in turbulent markets and when leadership within the index is narrow. RSP's gap to SPY was widest in 2009 and in 2023, the latter a year in which a handful of very large companies drove the index.
- The information ratio ignores beta. It credits a fund for any lead, including one that comes from taking more market risk. Alpha: the return that market exposure does not explain adjusts for that; the information ratio does not unless the benchmark is matched to the fund's exposure.
Compared with its neighbours
| Measure | Return in the numerator | Risk in the denominator | Alternative it assumes |
|---|---|---|---|
| Sharpe ratio | Return above the risk-free rate | Volatility of that excess | Cash |
| Information ratio | Return above a benchmark | Tracking error | An index fund |
| Alpha | Return beyond beta × market excess | (none; a return, not a ratio) | A levered index fund |
| Appraisal ratio | Alpha | Volatility of the regression's leftovers | A levered index fund |
| Tracking difference | Average return minus the index's | (none) | The index itself |
How Finabase does it
Finabase does not calculate tracking error or an information ratio. Nor does it assign a benchmark to a portfolio: the portfolio history chart plots the portfolio's own value with no comparison line.
What it provides is the visual version of the first figure in this lesson. On a symbol's comparison tab, adding SPY (or any other fund) redraws both as percent changes from the first bar in view. Lines that stay together have a low tracking error. Lines that fan apart have a high one, and the distance between them at the right-hand edge is the cumulative active return over the range. The table underneath gives each symbol's price return over periods from one week to two years, so the two-year active return is one subtraction away. Those figures exclude dividends, which matters when comparing funds with different yields.
For a number, export month-end prices of the fund and the benchmark, compute each month's
percent change, subtract, and apply STDEV.S(...) * SQRT(12) to the differences.
Frequently asked questions
What is a good information ratio?
Among professional managers, a ratio of 0.5 sustained over many years is regarded as very good and 1.0 as exceptional. Over three to five years, ratios that high occur often by chance.
What is a normal tracking error?
Index funds: a few hundredths to a few tenths of a percent. Funds that reweight an index (value, equal weight, low volatility): roughly 4% to 8%. Concentrated active funds: 10% and above.
Is tracking error the same as tracking difference?
No. The difference is the average shortfall against the index, mostly costs. The error is how much that shortfall varies. An index fund can have a low error and a meaningful difference.
Can the information ratio be negative?
Yes, whenever the fund trailed its benchmark on average. VTV's was −0.21 against SPY over 2005 to 2026.
Which benchmark should be used?
The index the fund is trying to beat or track, or the cheapest index fund that holds the same kind of assets. A benchmark the fund does not resemble inflates tracking error and makes the information ratio a comparison of two markets.
Summary
Active return is the fund's return minus the benchmark's. Its standard deviation, annualized, is the tracking error, and its average divided by the tracking error is the information ratio. Against SPY, an S&P 500 fund has a tracking error near 0.4%, reweighted versions of the index about 5.5%, the Nasdaq-100 8% and a concentrated active fund 27.5%. Tracking error tells an investor how far from the index to expect to be in a normal year, in either direction. The information ratio says whether the distance was rewarded, and it takes one or two decades of data before the answer is more than noise.
Check your understanding
5 questions on this lesson. Each answer comes with the reason, and nothing is scored.
A fund has a tracking error of 6% against its index. In one year it trails the index by 8 points. How should that be read?
- Something has gone wrong with the fund
- It is within the range to expect: gaps beyond one tracking error occur about one year in three
- The tracking error was calculated wrongly
Show the answer
B. It is within the range to expect: gaps beyond one tracking error occur about one year in three Tracking error is a standard deviation. Roughly two years in three fall within one tracking error of the benchmark and one in three outside it, in one direction or the other. Eight points against 6% is ordinary.
Fund A beats its benchmark by 2% a year with a tracking error of 4%. Fund B beats it by 3% with a tracking error of 12%. Which has the higher information ratio?
- Fund A: 2 ÷ 4 = 0.5 against 3 ÷ 12 = 0.25
- Fund B, because it earned more
- They are equal
Show the answer
A. Fund A: 2 ÷ 4 = 0.5 against 3 ÷ 12 = 0.25 The information ratio is the average active return divided by the tracking error. Fund A earned more lead per unit of deviation from the benchmark.
An S&P 500 index fund has a tracking error of 0.05% and trails its index by 0.4% every year. What do the two numbers say?
- It replicates the index closely, and its costs take about 0.4% a year
- It is a poor replication of the index
- It is nearly riskless
Show the answer
A. It replicates the index closely, and its costs take about 0.4% a year Tracking error measures the variability of the gap and tracking difference its average. A steady shortfall means good replication with a cost. Neither says anything about the index’s own risk.
A manager shows an information ratio of 0.6 over four years. Roughly how strong is the evidence?
- Strong: 0.6 is an excellent ratio
- Weak: 0.6 × √4 = 1.2, below the level of about 2 that separates skill from chance
- None can be drawn without the Sharpe ratio
Show the answer
B. Weak: 0.6 × √4 = 1.2, below the level of about 2 that separates skill from chance Multiplying the information ratio by the square root of the number of years approximates the t-statistic. At 1.2 the result is well within what luck produces; a true ratio of 0.5 needs about 16 years to reach 2.
A small-company fund shows a tracking error of 9.6% against the S&P 500. What does that mostly reflect?
- The manager’s stock picking
- The difference between small and large companies: the benchmark does not match the fund
- High fees
Show the answer
B. The difference between small and large companies: the benchmark does not match the fund A benchmark the fund does not resemble produces a large tracking error by construction. Against a small-company index the same fund would show a small one.
Sources and further reading
- Richard C. Grinold and Ronald N. Kahn, Active Portfolio Management, 2nd ed. (McGraw-Hill, 2000) — the information ratio and the fundamental law of active management
- Thomas H. Goodwin, “The Information Ratio,” Financial Analysts Journal, vol. 54, no. 4 (1998)
- U.S. Securities and Exchange Commission, Investor.gov — Index Funds (tracking error; underperformance from fees, trading costs and tracking error)
- Finabase chart data — monthly bars and dividends of IVV, RSP, VTV, QQQ, IWM, USMV, ARKK and SPY through October 2026; statistics computed by Finabase
