Intuition The big pictureWhen you buy a fund or ETF, you are hiring a manager to hold a basket of stocks for you. Two silent forces decide whether you actually get what the index earned:
The expense ratio — the yearly fee the fund quietly deducts from your money.
The tracking error — how much the fund's return wobbles away from the index it promises to copy.
A fund can advertise "we follow the NIFTY 50" but still hand you less. These two numbers tell you how much less and how reliably .
The expense ratio (ER) is the fund's total annual operating cost expressed as a percentage of the fund's assets:
ER = Total annual fund costs Average assets under management (AUM) \text{ER} = \frac{\text{Total annual fund costs}}{\text{Average assets under management (AUM)}} ER = Average assets under management (AUM) Total annual fund costs already subtracted from the NAV daily — you never get a separate bill.
WHY it exists: running a fund costs real money (salaries, exchanges, compliance). The fund recovers this by shaving a tiny slice off assets every day.
WHAT it means for you: if the index returns 12% and the ER is 1%, your gross-to-net leak is roughly 1% every single year , forever.
HOW it's charged: not once a year in a lump — it's accrued daily . A 1% annual ER means about 1 % 365 ≈ 0.00274 % \frac{1\%}{365}\approx 0.00274\% 365 1% ≈ 0.00274%
Suppose the underlying basket grows at rate r r r f f f P 0 P_0 P 0
Gross value after 1 year (no fee): P 0 ( 1 + r ) P_0(1+r) P 0 ( 1 + r ) P 1 = P 0 ( 1 + r ) ( 1 − f ) P_1 = P_0(1+r)(1-f) P 1 = P 0 ( 1 + r ) ( 1 − f )
After n n n P n = P 0 ( 1 + r ) n ( 1 − f ) n P_n = P_0\,(1+r)^n\,(1-f)^n P n = P 0 ( 1 + r ) n ( 1 − f ) n
Definition Tracking Difference vs Tracking Error
Tracking difference (TD): the average gap between fund return and index return over a period. TD = R fund − R index ‾ \text{TD} = \overline{R_{\text{fund}} - R_{\text{index}}} TD = R fund − R index bias (usually slightly negative because of fees).Tracking error (TE): the standard deviation of the return differences . It measures consistency , not size, of the gap.
TE = 1 N − 1 ∑ i = 1 N ( d i − d ˉ ) 2 , d i = R fund , i − R index , i \text{TE} = \sqrt{\frac{1}{N-1}\sum_{i=1}^{N}\big(d_i - \bar d\big)^2},\quad d_i = R_{\text{fund},i} - R_{\text{index},i} TE = N − 1 1 ∑ i = 1 N ( d i − d ˉ ) 2 , d i = R fund , i − R index , i
Let d i d_i d i i i i d ˉ \bar d d ˉ variance of the differences is the average squared deviation:
Var ( d ) = 1 N − 1 ∑ ( d i − d ˉ ) 2 \text{Var}(d) = \frac{1}{N-1}\sum (d_i - \bar d)^2 Var ( d ) = N − 1 1 ∑ ( d i − d ˉ ) 2 N − 1 N-1 N − 1 d ˉ \bar d d ˉ TE = Var ( d ) \text{TE} = \sqrt{\text{Var}(d)} TE = Var ( d )
WHY standard deviation and not just the average gap? Two funds could both average –0.5% vs index. Fund A does it smoothly (TE tiny); Fund B swings +3%, –4%, +2%... (TE huge). Fund B is unpredictable and hence riskier to hold as a "tracker," even if the average looks the same.
Cause
Why it creates a gap
Expense ratio
Fees drag return below index every day
Cash drag
Un-invested cash (for redemptions) doesn't rise with market
Sampling
Fund holds a subset of index stocks, not all
Rebalancing / index changes
Buying/selling at prices ≠ index snapshot
Dividend timing
Fund receives/reinvests dividends on different dates
Securities lending income
Can reduce tracking difference (adds return back)
Worked example Example 1 — Fee drag over 30 yearsIndex returns r = 10 % r=10\% r = 10% 0.1 % 0.1\% 0.1% 1.5 % 1.5\% 1.5%
Step 1 — cheap fund net multiple.
( 1 + 0.10 ) 30 ( 1 − 0.001 ) 30 (1+0.10)^{30}(1-0.001)^{30} ( 1 + 0.10 ) 30 ( 1 − 0.001 ) 30 Why? Each year grows at 10% then loses 0.1% of assets.
( 1.10 ) 30 = 17.449 (1.10)^{30}=17.449 ( 1.10 ) 30 = 17.449 ( 0.999 ) 30 = 0.9704 (0.999)^{30}=0.9704 ( 0.999 ) 30 = 0.9704 = 16.93 =16.93 = 16.93 ₹16.93 lakh .
Step 2 — expensive fund net multiple.
( 1.10 ) 30 ( 1 − 0.015 ) 30 = 17.449 × ( 0.985 ) 30 = 17.449 × 0.6344 = 11.07 (1.10)^{30}(1-0.015)^{30}=17.449\times(0.985)^{30}=17.449\times0.6344=11.07 ( 1.10 ) 30 ( 1 − 0.015 ) 30 = 17.449 × ( 0.985 ) 30 = 17.449 × 0.6344 = 11.07 ₹11.07 lakh .
Step 3 — the gap. ₹16.93L – ₹11.07L = ₹5.86 lakh lost to a "small" 1.4% extra fee.
Why so large? Because ( 1 − f ) n (1-f)^n ( 1 − f ) n
Worked example Example 2 — Computing tracking errorMonthly return differences d d d − 0.10 , − 0.05 , − 0.20 , 0.00 , − 0.15 -0.10, -0.05, -0.20, 0.00, -0.15 − 0.10 , − 0.05 , − 0.20 , 0.00 , − 0.15
Step 1 — mean. d ˉ = − 0.10 − 0.05 − 0.20 + 0.00 − 0.15 5 = − 0.50 5 = − 0.10 % \bar d = \frac{-0.10-0.05-0.20+0.00-0.15}{5} = \frac{-0.50}{5} = -0.10\% d ˉ = 5 − 0.10 − 0.05 − 0.20 + 0.00 − 0.15 = 5 − 0.50 = − 0.10% Why? This is the tracking difference (the bias) ≈ the fund's fee drag.
Step 2 — deviations from mean. 0.00 , + 0.05 , − 0.10 , + 0.10 , − 0.05 0.00, +0.05, -0.10, +0.10, -0.05 0.00 , + 0.05 , − 0.10 , + 0.10 , − 0.05 Why? We want spread around the bias, not around zero.
Step 3 — squared deviations. 0 , 0.0025 , 0.01 , 0.01 , 0.0025 0, 0.0025, 0.01, 0.01, 0.0025 0 , 0.0025 , 0.01 , 0.01 , 0.0025 = 0.025 =0.025 = 0.025
Step 4 — variance with N − 1 = 4 N-1=4 N − 1 = 4 0.025 / 4 = 0.00625 0.025/4 = 0.00625 0.025/4 = 0.00625 Why N − 1 N-1 N − 1 We used the sample mean, losing one degree of freedom.
Step 5 — TE (monthly). 0.00625 = 0.079 % \sqrt{0.00625}=0.079\% 0.00625 = 0.079%
Step 6 — annualize. TE annual = 0.079 % × 12 = 0.27 % \text{TE}_{\text{annual}} = 0.079\%\times\sqrt{12} = 0.27\% TE annual = 0.079% × 12 = 0.27% Why 12 \sqrt{12} 12 Variance adds over independent months; std-dev scales with time \sqrt{\text{time}} time
Worked example Example 3 — Two funds, same difference, different errorFund A differences: − 0.5 , − 0.5 , − 0.5 -0.5,-0.5,-0.5 − 0.5 , − 0.5 , − 0.5 + 2.0 , − 3.5 , − 0.5 +2.0,-3.5,-0.5 + 2.0 , − 3.5 , − 0.5 = − 0.5 % =-0.5\% = − 0.5% = 0 =0 = 0 Conclusion: A is a better tracker despite identical average gap. Same TD, very different reliability.
Common mistake "A lower expense ratio always means a better tracker."Why it feels right: fees are the biggest steady cause of underperformance, so cheaper seems strictly better.
The fix: ER is only the bias . A fund with 0.05% ER but poor sampling can have higher tracking error than a 0.10% ER fund. Check both TD and TE together.
Common mistake "Tracking error tells me how much the fund lost vs the index."Why it feels right: the word "error" sounds like a loss amount.
The fix: Amount lost = tracking difference . Wobble/inconsistency = tracking error . TE can be big even if average performance perfectly matches the index.
Common mistake "1% ER is small, ignore it."Why it feels right: 1% sounds trivial next to 10% market returns.
The fix: Over 30 years ( 1 − 0.01 ) 30 ≈ 0.74 (1-0.01)^{30}\approx 0.74 ( 1 − 0.01 ) 30 ≈ 0.74 edge . Compounding makes small fees enormous.
Common mistake "Tracking error can be negative."Why it feels right: confusing it with tracking difference, which is often negative.
The fix: TE is a standard deviation → TE ≥ 0 \text{TE}\ge 0 TE ≥ 0
What is the expense ratio? Annual fund operating cost as a % of AUM, deducted daily from NAV.
Is the expense ratio billed separately to investors? No — it's silently subtracted from the fund's NAV every day.
Net value formula for fund after n years with return r and fee f? P 0 ( 1 + r ) n ( 1 − f ) n P_0(1+r)^n(1-f)^n P 0 ( 1 + r ) n ( 1 − f ) n .
Fraction of wealth lost to fees over n years vs a zero-fee fund? 1 − ( 1 − f ) n 1-(1-f)^n 1 − ( 1 − f ) n .
Define tracking difference. The average gap between fund return and index return (the bias, usually ≈ –ER).
Define tracking error. The standard deviation of the fund-minus-index return differences (consistency of tracking).
Formula for tracking error? 1 N − 1 ∑ ( d i − d ˉ ) 2 \sqrt{\frac{1}{N-1}\sum(d_i-\bar d)^2} N − 1 1 ∑ ( d i − d ˉ ) 2 where
d i = R f u n d , i − R i n d e x , i d_i=R_{fund,i}-R_{index,i} d i = R f u n d , i − R in d e x , i .
Why use N–1 in tracking error? Bessel's correction — the sample mean was estimated from the data, using one degree of freedom.
Can tracking error be negative? No, it's a standard deviation, so ≥ 0. Only tracking difference has a sign.
How do you annualize a monthly tracking error? Multiply by
12 \sqrt{12} 12 (std-dev scales with √time).
Name three causes of tracking error besides fees. Cash drag, sampling (partial holdings), rebalancing/dividend-timing.
Two funds have identical tracking difference; which is the better tracker? The one with the lower tracking error (more consistent).
Why do small expense ratios matter over long horizons? Fees compound:
( 1 − f ) n (1-f)^n ( 1 − f ) n eats future growth, so a 1% fee can cost ~26% of returns over 30 years.
Which quantity approximates the fund's fee drag? The tracking difference (≈ negative of the expense ratio).
Recall Feynman: explain to a 12-year-oldImagine you copy your friend's homework (the friend is the "index"). The expense ratio is like your pen leaking a little ink on every page — you always end up with slightly less than your friend, and the more pages you do, the more ink you lose. The tracking error is how shaky your handwriting is: even if you copy the same average amount, sometimes you write way more, sometimes way less. A good copier loses very little ink (low fee) and has steady handwriting (low tracking error).
"Difference is the DIP, Error is the JITTER."
D ifference → D irection (has a sign, ≈ –fee, the average dip below index).E rror → variability (jitter, always ≥ 0).
And for fees: "Small fee, slow bleed." ( 1 − f ) n (1-f)^n ( 1 − f ) n
Intuition The 80/20 takeaway80% of long-run underperformance in index funds comes from two levers : keep the expense ratio low (controls the steady bias) and demand low tracking error (controls reliability). Master these two and you've mastered most of fund selection.
Index Funds vs Active Funds — why ER differences drive the debate.ETFs and NAV vs Market Price — premium/discount adds to tracking error.Compounding and Time Value of Money — why ( 1 − f ) n (1-f)^n ( 1 − f ) n Standard Deviation and Variance — the math engine of tracking error.Total Expense Ratio (TER) Regulations — caps and disclosure rules.Sampling vs Full Replication — a structural source of tracking error.
Loss fraction 1 minus 1-f to n
Intuition Hinglish mein samjhoDekho, jab aap koi index fund ya ETF kharidte ho, to do cheezein chupke se aapka return kha jaati hain. Pehli hai expense ratio — yeh fund ki saalana fees hai jo aapke paison me se roz thodi-thodi kaat li jaati hai. Aapko koi alag bill nahi aata, NAV me se hi cut ho jaata hai. Yeh 1% chhota lagta hai, par 30 saal me ( 1 − f ) n (1-f)^n ( 1 − f ) n
Doosri cheez hai tracking error . Yaad rakho, do alag alag terms hain: tracking difference matlab fund aur index ke return ka average farak (usually thoda negative, kyunki fees lag rahi hai). Aur tracking error matlab us farak ka standard deviation — yaani fund kitna consistent hai index ko copy karne me. Ek fund average me thoda peeche reh sakta hai par smooth (low TE), doosra average same par bahut ude-bhaage (high TE). High TE wala fund reliable tracker nahi hai.
Simple archer wali analogy: bullseye = index. Difference = aapke arrows ka center bullseye se kitna door hai. Error = arrows aapas me kitne bikhre hue hain. Achha index fund dono kam rakhta hai — center bullseye ke paas (fees ke barabar) aur arrows tight (kam wobble).
Isliye fund chunte waqt sirf sasta ER dekhna kaafi nahi — dono dekho: kam expense ratio (steady bias control) aur kam tracking error (reliability control). Yeh 80/20 rule hai fund selection ka: bas yeh do numbers master kar lo.
