Nigerian rate and currency premia reverse sign without an intercept

A zero-beta rate of 4.036% a month makes the macro premia in this paper hard to take as prices. Every one of the six portfolios it is supposed to price has a lower mean excess return. The highest is 2.209%. Owoputi, Akinyele, Akintan and Omorojor reach a null result on risk premia and acknowledge the small cross-section. They also flag how large both prices are relative to portfolio mean returns, and how much they depend on the fitted intercept. I would go further: this cross-section cannot identify a price for Nigerian rate or currency risk in either direction.

Six portfolios carry the pricing test

The paper uses macro APT in the Chen, Roll and Ross tradition. Unexpected rate increases raise discount rates and financing costs. A surprise depreciation of the naira helps firms earning foreign revenue and hurts importers. Portfolios that lose in states investors dislike should command a premium.

Each June, the authors sort NGX-listed firms into six value-weighted portfolios. Size splits at median June market cap; book-to-market splits at the 30th and 70th percentiles. The sample covers July 2016 to December 2025, giving 114 months of price returns without dividends. Both macro factors are AR(1) forecast residuals, measuring the unexpected part of each series. The rate factor uses the change in the 91-day Treasury-bill rate. The currency factor uses the log change in naira per dollar, with positive values denoting surprise depreciation.

The first pass estimates time-series betas on the macro factors alone. It then adds local CAPM, FF3, Carhart, FF5 and FF6 factors. FF6 adds momentum to FF5, whose profitability leg uses ROA. For the second pass, the authors run Fama-MacBeth regressions and report conventional, Newey-West and Shanken t-statistics. Their stated distinction is between "portfolio sensitivity from reliable compensation for bearing risk."

How much survives the equity controls?

All six interest-rate betas are negative in the macro-only model. Four clear 5%: S/L at -1.313 (t -2.796), B/L at -1.772 (t -2.034), B/M at -1.519 (t -3.213) and B/H at -1.464 (t -2.586). Every currency beta is positive. Only B/M is significant, at 0.305 (t 2.614). As the authors acknowledge, these are individual tests without a multiple-testing adjustment.

Most of the rate sensitivity disappears with controls. Adding the market on the 102-month common sample leaves only S/L near the significance threshold, at -0.894, t -1.969. With FF3, no interest-rate t exceeds 1.54 in absolute value.

The authors identify a limitation in that FF3 comparison: SMB and HML come from the same six test portfolios. The residual covariance consequently has rank four, making joint alpha tests unavailable. Adding macro factors raises mean R² from 0.617 to 0.621. Adjusted R² falls from 0.605 to 0.602, while mean absolute alpha increases from 0.168% to 0.173%. A benchmark built partly from the test assets was always going to leave little for the macro factors to add.

CAPM avoids that overlap. Here, mean R² rises from 0.340 to 0.348, but adjusted R² falls from 0.333 to 0.328. The interest-rate innovation has a -0.249 correlation with the market excess return. In CAPM plus macro, S/L alone retains an interest-rate loading near 5% (-0.894, t -1.969). The other five interest-rate t-statistics fall below 0.40 in absolute value.

A 3% monthly rate premium rests on the intercept

The extended-sample estimate prices interest-rate risk at 3.057% a month per unit beta, with a Newey-West t of 2.042. Its Shanken t is 0.882 (p 0.380), against a conventional Fama-MacBeth t of 1.908. Currency risk carries a price of 6.218% and a Shanken t of 0.666. The sharp drop under Shanken has a mechanical explanation: the correction increases with the price relative to factor volatility. At 3.057%, the rate price is almost twice the innovation's monthly standard deviation of 1.596%. The paper also acknowledges that a combined HAC-Shanken correction is not established.

The intercept does the heavy lifting.

As the authors observe, a positive interest-rate price paired with uniformly negative betas makes the fitted contribution of rate risk negative. The intercept supplies the level of returns. Remove it and both prices reverse sign. In the common-sample macro-only model, the interest-rate price shifts from +2.185% to -1.409%; the currency price shifts from +2.661% to -3.667%. That model has 3 cross-sectional degrees of freedom with an intercept. FF3 plus macro has 1 without one.

The conclusion describes the null as "insufficient evidence of a reliable premium, rather than proof that the risks have no economic relevance". It also says the sign reversals and small cross section constrain any interpretation of a stable premium. The risks may indeed matter economically. But I would press the remaining judgment further: estimates of this size establish nothing in either direction. A larger cross-section could change my mind, with industry or stock-level test assets yielding a premium and a cross-sectional intercept near zero. The authors' citation of Ang et al. points toward such a test.

Even the betas deserve caution given the inputs. Excess kurtosis reaches 31.396 for rate innovations and 22.404 for currency innovations. Rolling currency betas fall to about -41 in windows ending September 2019 to February 2020. The authors attribute this to nearly constant FX innovations. On 14 June 2023, the CBN adopted willing-buyer, willing-seller pricing. We did not find a subsample run excluding the unification months, nor did we find which naira rate the series uses. The authors cannot confirm whether they estimated the AR(1) recursively and treat the innovations as retrospective proxies.

Monitoring a Nigerian equity book

No premium survives the correction, leaving nothing here to put through costs or capacity. We also did not find firm counts for each portfolio. Even a surviving premium would therefore leave capacity unmeasurable. Among small firms, the descriptive value spread is 1.620 pp a month. This uses price returns, and S/H alone has an individually significant mean (t 2.549).

A gross stress estimate is the useful output. In the macro-only fit, B/M's beta of -1.519 translates a one-point surprise increase in the T-bill rate into roughly 1.5 pp of lost excess return that month. Once the market enters, the B/M interest-rate loading shrinks to -0.176 (t -0.386). The 1.5 pp estimate thus largely restates market beta. Rolling 36-month betas show S/H becoming positive around 2021 to 2022, while B/L turns more negative after 2023. The bands are pointwise and the windows overlap. They support frequent re-estimation, with no basis for a forecast.

We could not test this ourselves. Doing so requires NGX histories for prices, share counts and book equity, alongside Nigerian 91-day bill and naira series. We have none of those data. Substituting the US would test a different monetary and currency regime.

A desk can use the paper to monitor two macro exposures. It supplies no reliable evidence of what those exposures pay. The authors recommend precisely that use: track interest rate and currency exposures alongside conventional equity risks "without assuming that additional exposure generates higher expected returns."