Factor Investing vs Fundamental Investing Compared

Factor investing and fundamental investing differ in what defines the holding set. A factor approach sorts a universe by a stated rule and holds what the rule selects. A fundamental approach reaches a view on each business one at a time and holds what the view supports. Both end in a list of weights inside the same fixed sleeve, and both are judged by the same portfolio arithmetic.

Three figures sit still behind everything below. The equity sleeve of the Anantara Multi-Asset Portfolio is Rs 300 crore, fixed by the allocation work before a single name was considered. The mandate allows no holding above 5 per cent of the Rs 500 crore portfolio, or Rs 25,00,00,000/-. The limit was written before either approach was applied to anything. The holder is an invented charitable endowment whose committee Rukmini Deshpande chairs and whose mandate Faiz Ahmad Ansari runs.

Two ways of deciding what goes into that sleeve are set below against six stated criteria, and results are not one of the six. Neither approach is established as performing better than the other, and no evidence settles that question.

What actually differs between the two approaches?

One thing, and everything that follows rests on it. Factor investingFilling a sleeve by applying a stated rule to a stated list of candidates. A holding is there because the rule selected it, not because of anything specific to that holding. defines its holdings by applying a stated rule to a stated list of candidates. Fundamental investingFilling a sleeve by forming a view on each business separately. Every holding is there for a reason particular to it, not because of a rule applied across many. defines its holdings by forming a view on each business separately and funding the views that survive. The difference is in what defines the set, not in how much work goes into it. Either can be run with care or carelessly, so effort says nothing about which of the two is in view.

The everyday version. One household keeps a standing list and buys whatever has fallen below a quarter of a jar. The one next door buys what somebody in the kitchen thought worth buying that week. Both cupboards can hold the same rice and dal, and the difference shows next week rather than on today's shelf.

What is factor investing, stated on its own?

Four pieces, all four stated before anything is bought. First, a universeThe stated list of candidates a rule is allowed to consider. Anything outside it cannot be selected no matter how well it would have scored.: the list of candidates the rule may look at. Unlisted holdings are excluded, so the mandate has already narrowed this one and nothing unlisted can be selected however it scores. Second, a measurable characteristic every candidate in that universe can be given a number for. Third, a sorting ruleThe stated instruction that turns a ranked list of candidates into a selected set: rank on the measure, then take a stated portion or everything above a stated level.: rank the universe on that number and take a stated portion, or everything above a stated level. Fourth, a rebalancing scheduleThe stated dates on which a sort is run again and the sleeve is moved to whatever the fresh sort returns.: the dates on which the sort is run again and the sleeve is moved to the fresh answer.

All four have to be written down in advance, and it is that written-down quality rather than any particular characteristic that makes an approach a factor approach. The rule is stated before it is applied and applied afterwards without exception. Drop a name it returned and the sleeve stops being the output of the rule, so every claim made for the rule describes a set nobody holds.

The research tradition is named once and then left. Eugene Fama and Kenneth French, writing in 1993, set out a way of building a characteristic-sorted portfolio: rank a universe on a measure, form groups, study the difference. Which characteristics are used is covered under style factors later in this sequence; how return in a single portfolio is explained is covered under the factor model.

What is fundamental investing, stated on its own?

The other side, defined with the same care. A fundamental approach fills the sleeve one business at a time. Somebody forms a view about what each is worth against what it costs, and a view strong enough to fund becomes a holding sized to how strongly it is held. How a view is formed, researched and valued is settled under company research and valuation. A fundamental approach is fixed entirely by what defines its holding set, never by how the view inside it is reached.

Three portfolio properties follow, none of them about research. Each holding needs its own justification. Justifying one holding costs time that does not scale. The count is therefore bounded by how many views a team can hold at once, a capacity limit rather than a preference. Each view was argued on its own terms, so nothing obliges anybody to ask what the funded views have in common. The third property becomes the failure examined below.

The everyday version again. A street vendor stocking the stall reasons case by case, often excellently: this pickle sells on Fridays, that snack goes stale by noon, this drink moves when the office opposite is busy. Every decision is defensible. But ask what share of the stall rests on that one office and the method has no answer built into it. Counting is a different job from deciding.

What is each approach actually claiming?

Both sides are now defined, so the contrast can start. A factor claim is about a group: holdings sharing a stated characteristic behave differently, as a group, from holdings that do not. A fundamental claim is about a case: this business is worth more or less than its price suggests. The two claims are about different objects, and evidence for one is not evidence for the other.

Outside finance this is quick to feel. Ten shops in one mall share a driver of footfall, so their takings move together. The claim about shared footfall says nothing about whether the third shop from the left has a good manager, and an excellent manager there says nothing about the mall. Both can be true, and neither supports the other.

So each approach can be defended only with its own kind of evidence. A sorted group that separated over a long period defends the sort and never any single name inside it. A business that did what a view predicted defends that view, and one case is one case.

How many holdings does each end up with, and why?

Now the criterion where the arithmetic does the work. A rule returns however many names it returns, set by where the cut falls rather than by capacity: the top fifth of a wide universe may be several hundred. A set of views is bounded by how many a team can hold at once, and that bound is small because a view must be maintained, not merely formed. The bound on views is a capacity difference, not a conviction difference, and it produces two portfolios of completely different shape from the same Rs 300 crore.

Work it on the record. Rs 300 crore across 28 names gives an equally weighted holding of Rs 10,71,42,857/-, or Rs 10.71 crore. One holding is then 3.57 per cent of the sleeve and 2.14 per cent of the Rs 500 crore portfolio. Under a rule returning 200 names, Rs 300 crore over 200 is Rs 1,50,00,000/-, or Rs 1.50 crore. Each holding is 0.50 per cent of the sleeve and 0.30 per cent of the portfolio. Same money, same mandate, same sleeve, two portfolios that behave nothing like each other.

Watch the cap. At 28 names the mandate's Rs 25 crore limit is live: the record puts the largest holding at Rs 23,00,00,000/-, leaving Rs 2,00,00,000/- of room. At 200 names the average is Rs 1.50 crore, and a holding would have to be almost seventeen times that before the cap noticed it existed. The same sentence in the mandate is a binding constraint under one approach and completely inert under the other, and nobody changed a word of it.

Name the base whenever that Rs 23 crore is quoted. The largest holding is 4.6 per cent of the Rs 500 crore portfolio and 7.67 per cent of the Rs 300 crore sleeve. The cap is struck on the portfolio, and that base decides whether the limit is met. The sleeve base says how much of the equity work rests on one name. A committee that hears only the 4.6 has been told the truth and left with the wrong impression.

A floor hides underneath all of this, and it belongs to the mandate rather than to either approach. On an equally weighted sleeve no average holding may exceed Rs 25 crore, and Rs 300 crore divided by Rs 25 crore is 12. The mandate itself sets a minimum of twelve names, identically whether a rule or a set of views produced the list.

What does each approach cost to run?

The clearest practical difference is a cost difference rather than a quality one. A rule rebalancing on a schedule trades when the schedule arrives, changed situation or not. A view that has not changed generates no instruction at all and can sit for years. The scheduled approach pays for its discipline in turnover, and the view-led approach pays for its patience in the risk that nothing gets re-examined.

Turnover for the one stated twelve month period was 34 per cent of the Rs 500 crore portfolio. The chain built earlier in this sequence under trading cost is what both approaches meet: 34 per cent of Rs 500 crore is Rs 170 crore replaced; a replacement is a sale and a purchase, so Rs 340 crore went through the market; and Rs 340 crore against Rs 500 crore is a multiplier of 0.68. The drag is 0.68 times the cost rate, and no cost rate is supplied, so it stays an expression rather than a number.

The turnover chain contains no term describing who decided to trade or why. A rupee traded because a schedule arrived and a rupee traded because a view changed meet the same multiplier. The cost side of the comparison is trade frequency and nothing else.

How does somebody else check either claim?

A claim nobody outside the room can examine is not much use to the room. A factor claim is checked by reproducing the sort: the universe, the measure, the ranking and the cut were written down in advance, so an outsider can run those four statements over the same period and see whether they get the same holding setThe particular collection of holdings a sleeve ends up containing, considered as a set rather than as individual positions.. If they do not, one of the four statements was incomplete. Reproducibility is the factor approach's structural advantage, and it comes from having had to write everything down rather than from the rule being clever.

A fundamental claim is checked on a much smaller unit. The claim is about one business, so the test is what that business did next against what the view said, and the answer can come back clearly negative on a single case in a way a group claim almost never can. The factor claim is the more reproducible and the fundamental claim the more falsifiable case by case, and those are two different virtues rather than two scores on one scale. A committee wanting the first and one wanting the second are asking for different things, not disagreeing about which is better.

What would show that either had stopped working?

A claim that cannot fail is not a claim, so both approaches have to say in advance what would count as failing. FalsificationStating in advance what result would count as showing a claim wrong. The claim can then be tested rather than merely defended. here is the difference between a method and a story told afterwards.

For a factor claim the failure test is that the sorted groups stop separating over a window long enough that bad luck becomes implausible. The claim is statistical and short windows are noisy, so the separating window is long by construction. For a view on one business the test is that the business does the thing the view said it would not, and that can show inside a year. Both tests are slow relative to how quickly people lose patience, and that impatience produces most of the switching between the two.

A committee that switches after a poor stretch has not tested anything. The committee changed method before the old failure test had time to return an answer, then started a fresh long window from zero with the evidence about the old approach thrown away half collected.

Where do the two approaches meet the same constraints?

Both approaches hand over the same object, a list of names with a rupee weight against each, and that list goes through a mandate written before either method was applied to anything. The constraints were fixed before either method produced a single name, so the mandate cuts a factor rule and a set of separate views identically.

Make it concrete. A rule's cut puts Rs 32,00,00,000/- into one name; separately, a set of views wants Rs 32,00,00,000/- in one business. Rs 32 crore is 6.4 per cent of the Rs 500 crore portfolio and 10.67 per cent of the sleeve, against a cap of Rs 25,00,00,000/-. Both lists are cut to Rs 25 crore, and Rs 7,00,00,000/- goes elsewhere in both cases.

The convergence goes further than the cap. The Rs 300 crore sleeve, the ban on unlisted holdings and the cost rate on Rs 340 crore of traded value apply either way, and anything achieved inside the sleeve reaches the Rs 500 crore portfolio at 0.60 of its size. Choosing a method changes what the list contains and changes nothing about the container it is poured into.

Can a portfolio be both at once, and how would anyone tell?

Yes. A portfolio built from views alone carries a measurable loadingA number describing how strongly a portfolio moved with a stated common driver over a stated period, obtained by fitting rather than by asking anybody what they intended. on a common characteristic nobody selected for, and one built by rules can carry concentrations nobody chose. Measuring the exposures shows what a portfolio IS, and how it was assembled shows only how it was assembled. The two are routinely confused.

The household version. Twenty-eight decisions taken separately by people reading the same newspapers in the same month are not twenty-eight independent decisions. The twenty-eight are one decision with twenty-eight signatures on it. Nobody chose to bet the sleeve on a shared idea. Every conversation was the same conversation, so the idea survived all of them.

So can the Anantara portfolio be classified? No, and the missing measurement is itself the finding. The record carries one loading of 1.08 for the stated year, no characteristic measured across the 28 holdings, and nothing about the individual names. Reading a whole way of working out of a single number is the leap to refuse.

What do the six criteria look like set out together?

Everything above, on one grid. Read the columns as descriptions rather than as scores. Six criteria, twelve statements, and not a single verdict among them.

What does the year's own result add to this comparison?

Almost nothing. The portfolio returned 14.2 per cent gross for the stated twelve month period against 12.6 per cent for the composite benchmark, a gross excess of plus 1.6 percentage points, and that figure does not even survive its own cost line unchanged.

Take the fees off. Rs 6,25,00,000/- of management fee plus Rs 3,15,00,000/- of performance fee is Rs 9,40,00,000/-. On Rs 500 crore that is 1.88 per cent, so the net return is 12.32 per cent against the same 12.6 per cent benchmark. Plus 1.6 points of gross excess is a net shortfall of 0.28 points for the same year, and quoting either without saying which leaves the reader unable to tell whether the year went well.

The same Rs 6,25,00,000/- of management fee is also what an attribution selection effect of 1.25 points comes to on Rs 500 crore. Unrelated quantities, same digits, so name which one is meant every time.

The gross 1.6 also splits two incompatible ways. Attribution gives 0.35 of allocation and 1.25 of selection. The exposure split asks how much was simply carrying more market, and gives 0.488 and 1.112 of residual, from 6.5 plus 1.08 times 6.1, or 13.088 per cent expected, subtracted from 14.2. Both pairs sum to 1.6 and neither is the true split, so never take one term from each.

How does a committee actually use this comparison?

Rukmini Deshpande, chairing the endowment's investment committee, treats a proposal to change approach as a question about which of the six criteria is in play. One about trading frequency is a cost conversation, beside the 0.68 multiplier. One about outside checking is a reproducibility conversation. One about a poor stretch is neither, and naming that in the room is the most useful thing said that morning. Faiz Ahmad Ansari points the question at himself: what measurement would he offer if somebody disbelieved his description of the sleeve. Describing a portfolio by its process is a claim about the process and never a measurement of the portfolio.

An analyst runs the screen before opening any performance figure: which criteria are documented and which asserted, and can a rules based sort be re-run from what was published, or a view based record checked against what each view predicted. A household runs it without technical vocabulary: what would have to happen before this manager concluded their approach had stopped working, and how long would that take to show. Somebody who cannot answer the first has no failure test; a few months to the second is measuring noise.

When does the distinction stop mattering?

Four conditions close this question before anybody argues it. Under each, choosing a rule or choosing a set of separate views does not change what the holder ends up holding.

The first is breadth. Where the sorted set and the picked names overlap heavily, the two sleeves are one sleeve. Take the wide sort worked above, 200 names at Rs 1,50,00,000/- each. A 28 name book at Rs 10,71,42,857/- each whose names all sit inside those 200 differs in weights, not in what it holds, and both meet the same Rs 25 crore cap and the same 0.60 reach into the portfolio. At that overlap the method argument is deciding a few crore of weighting and nothing else.

The second is vocabulary. A set of views whose funding criteria are fixed and measurable is a sort written in prose. A committee that will fund only businesses cheap against their own accounts and quiet for a year has stated a universe, a measure and a cut without using those words, and lands close to the book a rule with the same content returns. The reasoning differs. The set does not, and the set is what is held.

The third is time. Both failure tests need a window, so a holder who will be out inside a quarter has bought neither thesis: neither has had time to express itself. Such a holder has bought the turnover instead, and turnover meets the same 0.68 multiplier whichever method generated it.

The fourth is the mandate. Nothing unlisted may be selected however it scores, no holding may exceed Rs 25,00,00,000/-, and an equally weighted Rs 300 crore sleeve cannot hold fewer than twelve names. Narrow the permitted candidates far enough and a rule and a set of views are choosing from the same short list.

None of the four announces itself when it stops applying. A universe widens at the next review, a cap is relaxed, a holder who expected to be out in a quarter is still there four years later. The condition that made the distinction irrelevant is gone, usually with no paper naming the moment, and the choice is live again inside a portfolio nobody has looked at that way since it stopped mattering.

The Anantara Multi-Asset Portfolio, the charitable endowment that holds it, its composite benchmark, Rukmini Deshpande and Faiz Ahmad Ansari are invented.
Educational material. Not advice on any investment, tax, budget or market position.