Class FamaMacBeth

java.lang.Object

com.quantfinlib.alpha.FamaMacBeth

public final class FamaMacBeth
extends Object

Fama-MacBeth (1973, J. Political Economy 81(3)) cross-sectional regression: what a factor exposure is worth per period in return space. Two passes:

1. each period, regress the cross-section of forward returns on that period's exposures (with intercept), giving one premium estimate λ_t per factor;
2. the premium is the time-series mean of the λ_t, with a time-series standard error. Each period contributes one observation per factor, which absorbs the cross-sectional correlation that wrecks a pooled regression's errors.

Standard errors are Newey-West (1987) on the λ_t series. With horizon-bar forward returns sampled every bar the λ_t overlap and are mechanically MA(horizon−1); the lag used is max(horizon − 1, Andrews 1991 AR(1) plug-in) per series. The iid t-stats are kept as naiveTStats for comparison. No Shanken (1992) errors-in-variables correction is applied: exposures here are observed characteristics (scores), not estimated betas, so the correction does not apply; pass estimated betas and it would.

Each cross-section is solved by Householder QR (Ols), not the normal equations. With standardize = true each exposure column is winsorised at mean ± 3 standard deviations and then z-scored per period, so one outlier cannot own the cross-section and premia are per unit of cross-sectional standard deviation.

NaN entries (asset not in that period's cross-section) are skipped per period; periods with fewer assets than factors + 2, or with a collinear design, are skipped and not counted. Infinite values throw.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final class	FamaMacBeth.Result

Method Summary

Modifier and Type	Method	Description
static FamaMacBeth.Result	fit(double[][][] exposures, double[][] forwardReturns)	One-bar forward returns, raw exposures.
static FamaMacBeth.Result	fit(double[][][] exposures, double[][] forwardReturns, int horizon, boolean standardize)	Fits the Fama-MacBeth (1973) two-pass cross-sectional regression.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

fit

public static FamaMacBeth.Result fit(double[][][] exposures,
 double[][] forwardReturns)

One-bar forward returns, raw exposures.

fit

public static FamaMacBeth.Result fit(double[][][] exposures,
 double[][] forwardReturns,
 int horizon,
 boolean standardize)

Fits the Fama-MacBeth (1973) two-pass cross-sectional regression.

Parameters:

exposures - exposures[t][asset][factor], the loadings known at t

forwardReturns - forwardReturns[t][asset], realised after t

horizon - bars spanned by each forward return; consecutive periods overlap by horizon − 1 bars when periods are one bar apart

standardize - winsorise and z-score exposures per period