Nigerian dividend yield's 'significant' result has a t-statistic of 0.36

This dividend-yield result gives a trader no basis for choosing one Nigerian consumer-goods stock over another. Every firm in a given year receives the same outcome value.

Okuku, Yakassai and Ene draw on the bird-in-hand view attributed to Gordon (1963) and Lintner (1962): investors value cash paid now more highly than retained earnings subject to uncertain reinvestment. Higher payouts should therefore accompany higher valuations, and greater retention should accompany lower ones. The authors examine all 21 consumer-goods firms listed on the Nigerian exchange as at 31 December 2024. Annual reports from 2014 to 2023 supply 210 firm-year observations.

Their multiple regression uses four payout and valuation ratios. Dividend per share, or DPS, enters in logs; dividend yield is DPS over price; earnings-to-price is EPS over price; and retention is (EPS minus DPS) over EPS. Return on assets is a control. The dependent variable is the natural log of the NGX consumer-goods sector index. Estimation is in EViews 12, with no panel effects reported.

The authors report a positive, significant yield result (p 0.0164), positive earnings-to-price at 0.061635 (p 0.0139), and negative retention at -0.06839 (p 0.0269). DPS is negative and insignificant at -0.014318 (p 0.197). The text gives R² as 0.30. The recommendations call for consistent payouts and lower retention.

What would the yield finding buy?

The Nigerian studies cited in the paper point in different directions. Kayode, Gbenga and Ayobami found a negative, significant yield effect among 20 listed firms over 2011 to 2020. Adegbie et al. found a positive, significant effect for insurers over 2008 to 2020. A sound positive result within consumer goods, paired with a retention penalty, would offer a tilt for a Lagos book. That is the paper's prospective contribution. Its tables do not support the yield leg.

Ten index prints, 210 rows

The sector index supplies one outcome per year. All 21 firms share that year's value, leaving ten distinct dependent-variable values across 210 observations. The index's standard deviation is 0.0516, and its range runs from 2.55 to 2.69.

A high-yield firm's index outcome is identical to its low-yield neighbour's. The regression can relate the firms' differing ratios to ten annual index prints, but it cannot measure which firm performed better. We did not find fixed effects, year effects or clustered errors in the paper. Without a reported correction, the standard errors treat firm-years as independent draws even though the outcome has just ten values.

The dividend-yield arithmetic

The reported yield slope is 0.056207 and its standard error is 0.154024. Their ratio is 0.3649, matching the reported t of 0.364923. Those three figures agree. The reported p of 0.0164 does not: a t of 0.36 implies a two-sided p of roughly 0.7 for any sample size relevant here.

The abstract opens its results with yield, saying it "exerts a positive and significant impact". In the conclusion, "Dividend Yield (DY) and Earnings to Price (EP) ratio have positive effects" immediately precedes the recommendations. Yield's pairwise correlation with the index is 0.174. The discussion attributes the disagreement with Kayode et al. to differences in sector and period. With this t-statistic, the paper has no yield finding to set against theirs.

Another conflict sits in the conclusion. It says Nigerian consumer-goods firms paying higher dividends tend to enjoy better market valuation, then says a few lines later that DPS does not significantly affect performance. The DPS slope is -0.014318; its correlation with the index is 0.012.

Which earnings measure?

E/P has the table's largest slope t, 2.53, but its label shifts throughout the paper. The descriptive table says EPS. The measurement table defines earnings per share divided by price. The regression table uses ESP (earnings to share price), while the abstract uses EP. Equation (1) names EPS and RP, neither of which appears in the estimated model. The methodology names DPS, DY and ESP as dividend-policy variables, omitting the retention ratio that is subsequently tested as a hypothesis.

The discussion treats the estimate as earnings per share and compares it with Oladimeji's (2023) EPS finding. Earnings in levels and earnings yield measure different things. Under the measurement table's definition, the variable is E/P: its mean is 0.2926, its maximum is 1.54, and its kurtosis is 10.67. A few extreme firm-years could carry considerable weight in that coefficient.

Yield and E/P both put price in the denominator. Their dependent variable is a price index from the same year, with no lags. The 0.061635 coefficient describes a same-year association rather than a forecast. The retention estimate is also hard to isolate: retention correlates -0.497 with DPS and 0.422 with E/P.

Constituent selection introduces a further timing concern. The sample starts with firms listed at the end of 2024 and follows them back to 2014, excluding any firms delisted along the way. Our FactorBench review identified the same fixed-constituent problem.

Nothing to size

We could not test the claim ourselves. We have no Nigerian equity prices, dividends or company fundamentals. Substituting US stocks would examine a different market and say nothing about the authors' claim. A direct check needs firm-level NGX price and dividend histories for these 21 names over 2014 to 2023.

The paper gives no returns, costs, holding rule or out-of-sample period, so there is nothing to size.

One result would change my view: firm-level returns regressed on lagged yield across these 21 stocks, with year effects absorbing the common index move, and a positive spread surviving Lagos transaction costs. Until then, the evidence consists of internally coherent E/P and retention coefficients (t 2.53 and -2.27) fitted to ten index values, alongside a yield coefficient whose t is 0.36.