Tracking error is often cited as a key factor in tracker fund selection. Tracking error is to tracker funds as goal tally is to a Premiership striker – a fundamental measure of how well the job is being done.
A tracker’s role is to deliver the returns of its benchmark index. In an ideal world, if the FTSE All-Share index returns 10% a year, then a FTSE All-Share tracker will also return 10%. But the world is rarely perfect (mine isn’t anyway), and tracking error shows just how wide of the (bench) mark the performance is.
That’s important information for passive investors because it can reveal the hidden cost of owning a tracker. There’s no point choosing a fund with a Total Expense Ratio (TER) 0.2% cheaper than its rivals, if its returns consistently lag the same benchmark by an extra 0.5%.
What exactly is tracking error?
Like so many investing terms, there are many different versions of tracking error and ways of calculating it. It’s therefore often difficult to be sure that two different sources are talking about the same measure.
That said, a reasonably common definition of tracking error is:
Tracking error = the standard deviation of returns relative to the returns of the index.
The lower the tracking error, the more faithfully the fund is matching its index.
When comparing funds, choose the lowest tracking error possible. A tracking error above 2% indicates the fund is doing a bad job.
What causes tracking error?
An index tracker is like an impressionist. It mimics its benchmark but can never quite be a dead ringer, chiefly because of:
- Costs
Index returns aren’t dragged down by operating expenses, but tracker fund returns are. Therefore you’d always expect a fund to lag its benchmark by at least its TER. The TER is deducted from the fund’s net asset value (NAV) on a daily basis, so the lower the TER, the lower the tracking error, and the better the expected fund return, all things being equal.
- Replication
Full replication funds that hold every stock in their index should offer zero tracking error as they are the index. (Except that transactions costs incurred when rebalancing ruin the beautiful dream).
Partial replication funds that sample a portion of their benchmark’s securities (because the cost of holding them all is too high) will inevitably generate tracking error, being only a representation of the index rather than a perfect clone.
Synthetic funds use swap-based contracts to guarantee they match their indices’ performance. But the swap fees and collateral costs incurred by the contracts drag down fund performance against the benchmark.
Taxes and transaction costs like brokerage fees and trading spreads add to our tracking error woes, no matter which replication model is used.
- Turnover
The more trades a fund makes, the greater the trading costs, which ultimately undermines performance and adds to tracking error.
- Management experience
Although index funds are popularly thought to be so simple that they can be run by VIC-20 computers, better management can rein in tracking error. Look for funds with a long record of tight tracking error.
- Enhancements
This one actually works in our favour. Funds can earn extra revenue that closes the performance gap (or even turns it positive). Typically this income comes from two sources:
- Fees earned by lending out fund securities to short-sellers.
- Dividend enhancement – lending out securities to tax-benign territories when dividends pay out.
Now what?
So that’s tracking error in a blog-style nutshell. And everything would be hunky dory if fund providers published tracking error on their factsheets just like TER. But they don’t, and there’s no regulatory requirement for them to do so.
Clean comparisons of funds by tracking error are nigh on impossible in the real world. You might work tracking error out for yourself, if you love equations, but that doesn’t sound like many couch-potato investors I know.
Instead, you can use tracking difference as a substitute for tracking error when you want to compare rival funds.
Take it steady,
The Accumulator
Comments on this entry are closed.
An aside…Vic-20, takes me back…! I was a Beeb man myself… After the good old ZX81 of course. Hard to argue with progress however… hard to imagine the Internet and those spongey keyboards.
1. As ever, thanks for this.
2. In your list of possible reasons for tracking error, you don’t consider the issue of percentage of assets in cash.
3. Also, fund size might be relevant. That is, larger funds may benefit from economies of scale.
4. Further, funds will differ in the number and size of redemption requests they receive over a given period of time.
I have become a fan of your blog. Translating the dank
imponderables of investing into one-byte intellectual snacks for
novices like me is not an easy task; but you do it very well. Apart
from the information contained in your posts I like your writing
style.
Interesting the issue on tracking error – I believe Tim Hale in his latest book(Smarter Investing) suggested that tracking error should not exceed 1% although I have to say having looked at a number of my HSBC index trackers most are over this figure and believe HSBC FTSE 250 under my Fidelity platform is showing a 2.95% tracking error – or am I missing something here!
To Steve: HSBC claim that the annualised 3 year tracking
error on their FTSE250 tracker is 0.55% on the product factsheet as
of 30th Nov 2010. Any ideas how Fidelity (or even HSBC, for that
matter) compute this? There should be an FSA requirement to publish
tracking error to an agreed/standardised methodology (as is being
half-attempted with TER) for all index tracking funds. Once TER
gets nailed down (I’m an optimist!) the next best way for index
funds to milk investors is by fat (unpublished) tracking errors.
One would hope the FSA would cotton onto this and lay down to the
law. And hope is just about all you have in this regard! Until
then, it’s really something of a guessing game and/or involves
significant effort. HSBC are one of the few who even publish
tracking errors at all – you’re right to question how accurate they
are. One crude way is just to compare the returns of the index vs
the tracker fund for whatever time interval suits you (x
months/years/decades). This gives you a quick “real world” idea of
tracker error but is rough-and-ready. As always, DYOR.
Rob – 100% agree it would seem that a large proportion of these companies are never very transparent.
I have to say Vanguard are one of the few exceptions and there is a comfort in knowing that the policy holders are also owners and their TER is what it implies – their total charges and not dressed up as something else.
@ Steve – you’ve put your finger on nub of the problem: we can’t compare tracking errors because we’ve no idea whether they’re comparable and because hardly anyone publishes them. During my research, I discovered HSBC’s FTSE 100 index fund (acc) awarded 0.38% tracking error by HSBC and 4.18% by TD Waterhouse (using Morningstar data). TDW’s data looks ludicrously high (and it’s roughly this level for every tracker I looked at on their site). I’ll get in touch with them and ask them how they calculate it.
Hale says keep tracking error low, ideally below 1% and that beyond 2% is unacceptable. But, it’s irrelevant as long as the measure remains so cryptic.
Rob is bang on with his reply.
I did a quick sweep of the major tracker fund players to see who publishes tracking error:
Yes
HSBC index funds
Credit Suisse ETF
No
iShares
db x-trackers
HSBC ETF
L&G
Vanguard
@ OldPro – I had a ZX Spectrum, though I enjoyed the odd game of Shadowfax on a friend’s VIC-20. Beeb mates provoked raging jealousy cos they had Elite. It somehow seemed rude to turn up at their house and expect to play on their computer for an 8-hour session – without talking -just so I could get to Mostly Harmless.
Fidelity’s data is taken from Morningstar, just like TD Waterhouse. One basic problem is they don’t say what timescale their version of tracking error is dealing with. Another problem is that some funds aren’t benchmarked against the right index. The FTSE 250 tracker is OK, but if you look at the HSBC FTSE All-Share fund, Morningstar are comparing it against the FTSE 100.
1. Hi TA, I notice you didn’t respond to my comments.
2. Can I get a refund? Ha ha.
Hi Alex,
I didn’t feel your comments needed discussion this time as I pretty much agree with everything you said. Although I’d expect cash holdings to have a miniscule effect in the average tracker. And that’s a no to the refund, I do credit notes only 😉
Very good explanation Monevator!
.-= Moneycone on: The Beardstown Ladies =-.
It seems to me that there are two aspects to the tracking error to consider:
the systematic error, causing a drift from the index over time
the random error, cause fluctuations about the index
The former would be due to trading costs and enhancements, and the latter caused by the replication error. Eg looking at HSBC FTSE All-Share tracker against its index over 12 months and 3 months on the H-L site, there seems to be quite a bit of random fluctuations, but they end up at pretty much the same place, suggesting low systematic drift. However, on the 6 month view they end up 2% different – and during December they were 4% different. So I guess over this particular period, the large fluctuations have failed to cancel out fully. (Perhaps TDW measured it in December ..?)
Is it systematic error we should be most worried about, or the size of the random fluctuations ? More accurate replication should cause smaller random fluctuations, but could increase costs resulting in a larger systematic drift.
Ultimately what I care about is that my returns match the index as closely as possible. Theoretically the fluctuations along the way are moot if we all get to the same place in the end. But the fluctuations tell their own story. If they’re wild then the tracker is probably not be the dead ringer for the index that it claims to be and that’s a warning sign that all may not be well in the long-term.
My next step would be to compare the fund with its nearest rivals, and if costs and tracking difference over time are in the same ball-park, then I’d choose the one that bucked around the least. Ultimately, I’d prioritise tracking difference but there’d need to be a reasonable track record to judge that by. I’d want at least 5 years, preferably 10 (I’ll be lucky), would settle for 3 (where I’ve got no choice, and I usually don’t).
Lending securities to short-sellers for a little extra return is not risk-free. Take another look at Nassim Taleb’s thoughts on options writers (“Fooled by Randomness”), and you’ll grok what I mean.
Indeed, there’s a whole dimension of tracking error which is not mentioned in this article: the risk profile of the tracker against that of theindex.
Still, a valuable article, and well-written. Thanks.