Seven Country-Years: What the ECOWAS Convergence Record Says About the Eco
By Dr. Boris Houenou ·
Seven out of one hundred and eighty
I hand-collected the compliance tables from ECOWAS convergence reports, because nobody publishes them in machine-readable form. The panel covers all fifteen member states from 2005 to 2016. That is 180 country-years of official assessment against the criteria that gate entry into the Eco.
Full compliance with every criterion in force occurs in seven of those 180 country-years. Just under four percent. All seven fall in the same year, 2016. And all seven are members of WAEMU: Benin, Burkina Faso, Côte d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal.
The countries that pass the test for joining a currency union are the countries that already share one.
That single fact should reorganize how the region thinks about the phased launch reaffirmed at the Lungi summit in July 2026. ECOWAS has now missed launch dates in 2003, 2005, 2009, 2015 and 2020. Each postponement carried the same diagnosis: too few countries met the criteria, so the launch must wait for convergence. The alternative inference, that a test failed by almost everyone for two decades might be the wrong test, has never been drawn.
The criteria do measure something real
Let me be fair to the framework before I take it apart. Compliance is not noise. In fixed-effects Poisson models that adjust for the number of criteria in force, one extra point of inflation reduces the number of criteria a country meets by 2.8 percent. Faster growth raises compliance. A stronger current account raises it. These are the signs you would want.
There is an obvious objection: the inflation ceiling is itself one of the criteria, so low inflation ticks a box by definition, not by behaviour. I tested it. Splitting the count into the primary block, which contains the inflation ceiling, and the secondary block, which does not, the entire inflation effect sits in the secondary block (coefficient of minus 0.041, standard error 0.009). The primary block, the one exposed to mechanical overlap, shows nothing (minus 0.014, standard error 0.022).
So inflation predicts compliance because it is a symptom of the fiscal and monetary configuration the other criteria score. The measurement is honest. The problem is elsewhere.
The region is diverging on the one metric surveillance was built to compress
Cross-country dispersion in the compliance share rises by 0.72 percentage points a year across the sample, with a p-value below 0.001. In plain terms, the standard deviation of compliance went from 0.13 in 2005 to 0.22 in 2016. Dispersion widened by roughly two thirds under two decades of multilateral surveillance. Growth dispersion shows no downward trend. Current account dispersion shows none. Inflation dispersion falls only weakly.
Individual countries do revert toward their own norms. They just do not revert toward each other. That combination, mean reversion within countries and widening spread across them, is the textbook signature of club convergence: a permanent compliant club and a permanent non-compliant club, not gradual accession.
A Markov analysis of compliance regimes puts numbers on the waiting time. A country starting in the low tercile takes an expected 7.8 years to reach the high tercile. From the middle, 6.8 years. That is longer than the interval between any two of ECOWAS’s five missed launch dates. And the high tercile only means meeting 60 percent of criteria, while the launch gate demands sustained full compliance, which appears in under four percent of observations. Those waiting times are floors, not estimates.
The test is nominal. The cost of a currency union is real.
Convergence criteria measure inflation, deficits, reserves, debt. What determines whether a shared currency hurts is whether members need the same monetary stance at the same time.
Across the 105 country pairs in ECOWAS, the mean correlation of annual real GDP growth over 2005 to 2016 is 0.00. Forty-three percent of pairs are negative. Nigeria, which accounts for roughly two thirds of regional output and would dominate any union, has a growth cycle essentially uncorrelated with the rest of the region, averaging minus 0.04.
Here is the part that should trouble anyone betting on a currency union to create its own optimality. Inside WAEMU, after decades of a shared currency, a supranational central bank and a common convergence pact, the mean pairwise correlation is 0.04. No better than the region at large. The mechanism that is supposed to synchronize members runs through trade intensity, and intra-ECOWAS trade has hovered around a tenth of members’ total merchandise trade throughout the convergence programme, broadly flat since the mid-1990s. A union cannot bootstrap itself through a channel that carries a tenth of its commerce.
Some of the recent progress is arithmetic
In 2015 the Authority rationalized the criteria set from eleven to six. Watch what that does to the headline. The mean share of criteria met jumps from 0.55 in 2014 to 0.66 in 2015 and 0.80 in 2016. Over the same period the mean count of criteria actually met falls from 6.0 to 3.9.
The denominator changed. The economies did not. And the surveillance reports do not flag the discontinuity when they present compliance trends.
The language converged while the numbers diverged
I ran the five convergence reports published for 2010, 2012, 2014, 2015 and 2016 through a battery of text-similarity measures. The last three form a tight cluster: Jaccard overlap of 0.450 to 0.461 and TF-IDF cosine similarity of 0.691 to 0.715, against 0.319 to 0.344 and 0.526 to 0.578 for pairings with the earlier reports. On a latent-semantic measure, the 2015 and 2016 reports sit at 0.963, close to an identical thematic mix.
The cheap explanation is copy-paste. It is wrong. Verbatim eight-gram overlap never exceeds 2.3 percent. And a permutation test that reshuffles paragraphs across reports produces documents that look more alike than the real ones: observed mean Jaccard of 0.375 against a null of 0.428, about fifteen standard deviations below. Each report carries genuine year-specific content.
Which means the cluster is deliberate. As the launch deadline approached, the institution consolidated its narrative. Over the same window, compliance dispersion hit its sample maximum. Policy language converged while policy outcomes pulled apart.
My own reading of this, beyond what the data shows, is that narrative alignment is simply the cheapest margin of integration. When fiscal adjustment and diversification stall, an institution facing a deadline can still coordinate its words. The capacity on display is real. It is currently pointed at the story rather than at the measurement.
What a serious 2027 looks like
Apply the record to the gate. A phased launch restricted to qualifiers produces, with high probability, the CFA franc zone plus one or two others. And three of the seven states that ever achieved full compliance, Mali, Burkina Faso and Niger, left ECOWAS in January 2025 while staying in WAEMU. The set of countries that are both in the launch pool and able to pass is now smaller than WAEMU itself.
Four things follow.
Name the selection outcome out loud and negotiate accession terms for Nigeria and Ghana before launch, not after. Founding configurations create incumbents, and every governance question left open by the 2019 to 2021 naming dispute gets harder once someone holds the seat.
Cyclically adjust the fiscal targets. Compliance collapsed in 2008, when the mean count of criteria met fell to 2.9. An unadjusted gate binds hardest in downturns and times the launch to global luck.
Build the insurance before the currency. A regional stabilization facility funded in good commodity years and disbursing on terms-of-trade triggers addresses the asymmetry the 2014 price collapse exposed. So does payments integration that lets regional trade settle without transiting third currencies.
And publish the compliance tables in machine-readable form, pre-register criteria definitions, and reconcile each year’s assessment with the last on a fixed definition. These cost a table and an afternoon. A currency’s credibility survives bad numbers honestly reported. It does not survive good numbers that turn out to have been definitional.
The convergence-first strategy has not failed for lack of time. It conditions monetary union on criteria that are endogenous to monetary union. Build the trade, the payments and the fiscal architecture first, and let the currency be the capstone rather than the foundation.