CFA® glossary: H

21 terms beginning with H.

ABCDEFGHIJKLMNOPQRSTUVWYZ0-9

H-model

A dividend model in which an above-normal growth rate declines in a straight line to a long-run rate, rather than dropping abruptly. It gives an approximate value in a single expression. Where it appears →

Hallucination

Fluent, confident model output that is false or unsupported, such as an invented figure or citation. It comes from generating plausible text, so it cannot be ruled out, only checked for.

Hazard rate

The probability of default in a short period given that no default has happened before it. Times the probability of surviving to the start of a period, it gives that period's default probability. Where it appears →

Headline inflation

Inflation measured across the whole basket, food and energy included. It is what households experience, and its most volatile components are the likeliest to reverse.

Hedge accounting

Accounting that lets gains and losses on a hedge be recognised in step with those on the exposure it offsets, so earnings do not swing with the hedge alone. The hedging relationship must be designated and documented.

Hedge ratio

The number of units of the underlying held per option to make the combined position riskless over the next period. In the binomial model, the change in option value over the change in the underlying's price. Where it appears →

Hegemony

Openness to global flows combined with non-cooperation: a dominant power sets the rules and exerts influence over others rather than negotiating with them as equals.

Herding

Following what other investors are doing in place of one's own analysis. For a professional it can be individually rational, since being wrong with the crowd is safer than being wrong alone. Where it appears →

Herfindahl-hirschman index

The sum of every firm's squared market share. Squaring weights large firms heavily, so it separates one dominant firm from an even split where a concentration ratio cannot.

Heteroskedasticity

Error variance that differs across observations, breaking the regression assumption of constant variance. It makes the reported standard errors unreliable, so significance tests can no longer be trusted. Where it appears →

High-water mark

A level a hedge fund must climb back above before charging an incentive fee again, so the same gains are not charged for twice after a loss. Where it appears →

Historical cost

Carrying something at what was originally paid. Verifiable, and capable of becoming badly stale on anything held for decades. Where it appears →

Historical simulation

Estimating a portfolio's possible outcomes by applying risk-factor changes actually observed over a past window to today's positions. It needs no distributional assumption but cannot show what the window never saw. Where it appears →

Holding period return

The total return over one period: price change plus income received, over the starting value. It assumes nothing about length or reinvestment, which makes it the building block for other return measures.

Homemade dividend

Cash an investor creates by selling some shares instead of receiving a dividend. With no taxes or costs it is equivalent to a dividend, which is why payout policy cannot change value in that world.

Hot money

Short-term capital that crosses borders chasing yield or sentiment and leaves as quickly as it came. Its sudden inflows and reversals destabilise asset prices and exchange rates. Where it appears →

Human capital

The value of the skills and knowledge people bring to work. For an individual, the present value of expected future earnings: largest early in a career, and bond-like or equity-like depending on how secure the income is. Where it appears →

Hurdle rate

A return a fund must clear before any incentive fee is earned. Where it appears →

Hyperinflation

Inflation so high, around 100% cumulatively over three years, that a currency stops being a stable measure. Financial statements are then restated for inflation, or remeasured, before translation.

Hyperparameter

A setting chosen before a model is trained, such as the size of a penalty or the depth of a tree, rather than estimated from the data. Usually tuned by cross-validation.

Hypothesis test

A procedure for judging whether sample evidence is strong enough to reject a stated claim about a population parameter, at a significance level chosen in advance.