An index fund is designed to deliver whatever the index delivers. It never quite does, and the size and direction of the difference is a measurable property of the fund rather than an accident.
The index is not an investable thing
An index is a calculation. It assumes positions are held in exact proportion, that dividends are reinvested instantly, and that no transaction costs exist.
A fund has to buy real securities in real markets at real prices, and every one of those assumptions fails slightly in practice.
The gap between the calculation and the achievable portfolio is the structural floor beneath which tracking cannot improve, regardless of how well the fund is run.
Fees are the largest predictable component
The ongoing charge is deducted from fund assets continuously, so a fund holding exactly the index will lag it by roughly the fee over time.
This is the one component that is known in advance and does not vary, which is why fee comparison explains most of the long-term difference between similar funds.
It also compounds. A small annual difference applied over decades produces a divergence far larger than the headline figure suggests.
Cash and timing create drag
Money arriving from new investors is not invested instantaneously, and dividends received from holdings sit as cash until they are deployed.
In a rising market that uninvested cash lags, and in a falling market it helps, which means the effect is not a consistent cost but a source of variability.
Funds manage this with futures and careful scheduling, which reduces the drag while introducing its own small costs and basis differences.
Sampling replaces full replication
Broad indices contain holdings that are illiquid or expensive to buy in small amounts, so many funds hold a representative subset rather than everything.
Sampling reduces trading costs and introduces the possibility that the subset behaves differently from the whole, particularly during periods when smaller holdings diverge from larger ones.
The trade-off is deliberate. A fund that replicated an index exactly would incur costs that damage returns more than the sampling error does.
Index changes force trading at bad prices
When an index adds or removes a constituent, every tracking fund must adjust on the same date, and the market knows this in advance.
That concentration of demand moves prices before the funds trade, so replication is achieved at prices worse than the index assumes.
Providers respond by spreading trades or using flexibility around the effective date, which improves outcomes while moving the fund slightly away from pure replication.