Owning the MineIs Not Taxing It: Burkina Faso and SOPAMIB

Abstract

Burkina Faso, Africa's fourth-largest gold producer, with output above 94 tonnes in 2025 on official figures, has chosen ownership over taxation. Under its 2024 mining code it created the Société de Participation Minière du Burkina (SOPAMIB), which by June 2025 had taken over two operating mines and three exploration licences from Endeavour Mining and Lilium, and the Prime Minister has said more nationalisations will follow. This paper takes the policy as given and asks the revenue question it leaves open: how a state that owns mines collects from them. A state-owned mine pays royalty, tax and dividends only if someone independent of its management verifies production and cost, and the state's revenue from a nationalised mine is a net figure after the capital it must now supply. The paper specifies a state-enterprise fiscal model that separates owner returns from tax, an ownership-versus-taxation comparison across price and cost scenarios, a physical-balance audit that the Treasury can run on SOPAMIB itself, and an artisanal channel model for the tens of tonnes that never reach any counter. The illustrative result is that at current prices ownership out-earns taxation only if the state's operating cost stays within about 10% of the private operator's, and that every percentage point of cost drift above that hands back more than a point of take.

Introduction

Mali renegotiated; Burkina Faso took the keys. The difference matters for revenue design. A private operator pays the state through instruments the state sets and audits. A state operator pays the state what is left after its own decisions on cost, capital and dividends, and the auditor and the audited report to the same minister.

The research question is how the Treasury, as distinct from the mining ministry and SOPAMIB, secures a predictable revenue stream from mines the state now owns, and how it decides which further mines are worth owning. The paper is written for the Ministry of Economy and Finance, the Cour des comptes and Parliament's budget committee, the three bodies that must be able to check SOPAMIB from outside.

The organising identity, adapted for a state operator:

R^{pot} - R = \underbrace{\left(R^{pot} - R^{law}\right)}_{\text{policy gap: dividend policy, capital calls}} + \underbrace{\left(R^{law} - R\right)}_{\text{compliance gap: unverified production and cost, artisanal leakage}}

Current tax issues

2.1 Owner versus tax collector

Burkina Faso's approach differs from Mali's in kind, not degree. Mali kept private operators and raised what it charges them. Burkina Faso, through the 2024 mining code and the creation of SOPAMIB, has taken operating control of mines: Wahgnion and Boungou from Endeavour Mining and Lilium in 2025, a third mine since, and a stated intention to acquire more. It has also taken a 15% stake in Essakane, the country's largest mine, and six of fifteen industrial mines are now majority Burkinabè-owned. The policy is framed as sovereignty; this paper takes it as given and asks what it does to revenue.

Under private ownership the state's revenue from a mine is a set of charges it defines and can audit: royalty on production, corporate tax on profit, dividends on its minority stake. Under state ownership the royalty and corporate tax become transfers from one arm of the state to another. They still appear in the mining ministry's revenue statistics, and they may still be paid into the Treasury, but they are not new money; they are the state paying itself out of a mine it now owns. The new money is the dividend, and the dividend is what remains after SOPAMIB has paid its own operating costs, financed its own capital, serviced any debt and decided how much to retain.

The revenue question is therefore whether SOPAMIB's dividend exceeds what a private operator would have paid in royalty, tax and minority dividends on the same mine, after the state has paid to acquire and restart it. That depends on three things the state controls and one it may not. It controls its own cost of capital, which is high while Burkina Faso is outside concessional markets; its dividend policy, which does not yet exist in published form; and the acquisition price, which for Wahgnion and Boungou was US$60 million plus a 3% royalty on 400,000 ounces. It may not control SOPAMIB's operating cost, which depends on whether a state company can run a mine as efficiently as Endeavour did, under worse security, with less access to spares, contractors and finance. Section 3.1 makes the comparison a formula and Part II runs it on the actual terms.

2.2 Capital and operating risk moves to the budget

A private operator is a risk absorber. When gold falls, when a pit wall fails, when a contractor's convoy is attacked or when the mill needs a new crusher, the operator's shareholders bear the cost and the state's royalty falls at most in proportion to output. When the state is the operator, every one of those events becomes a budget item. A price fall of 30% at a mine with an all-in cost of US$1,500 and a price of US$4,000 cuts the margin from US$2,500 to US$1,300 an ounce, nearly halving the dividend; a fall to US$2,000 cuts it to US$500 and, with any cost drift, to nothing. A mine that paid taxes under private ownership then requires a capital call from the Treasury to stay open, or closes and pays nothing.

Burkina Faso's budget is already gold-dependent. Mining revenue reached CFA 776 billion in 2025, and the IMF attributes the fiscal deficit's staying within the 4% of GDP programme objective largely to gold. Adding direct operating exposure at three mines to that dependence raises the variance of budget revenue in a way that the mining ministry's revenue statistics, which record royalty and tax as if the mines were still private, do not show. Section 3.4 sizes the exposure as a distribution and specifies a buffer; Part II gives the numbers.

2.3 Security costs

The mines Burkina Faso has taken over are in its most exposed regions. Boungou, in the east near the Niger border, was the site of a November 2019 attack on an employee convoy that killed dozens of people and closed the mine for months; Endeavour's willingness to sell it reflected the security cost as much as the ore body. Wahgnion, in the south-west near Côte d'Ivoire, has been less affected but is not immune. Under private ownership the security cost was borne by the operator through private security contracts, insurance premiums, convoy escorts and lost production. Under SOPAMIB it is a state cost, and the state is also the party providing the security, so the cost is partly hidden inside the defence budget.

For the models this means two things. The ownership-versus-taxation comparison in section 3.1 must include a security cost and a security-closure probability, both of which fall on the state under ownership and on the operator under taxation. And the buffer in section 3.4 must cover a scenario in which one state-owned mine is closed for a year by attack, which at Boungou has already happened once.

2.4 Verification of state production

When a private operator declares production, the mining ministry and the tax authority have an incentive to check it, because the state's royalty and tax depend on the figure and the operator's incentive runs the other way. When SOPAMIB declares production, the mining ministry that supervises SOPAMIB, the board that SOPAMIB reports to and the Treasury that receives its dividend all sit on the same side of the table. The operator and the auditor report, in the end, to the same minister. Ecofin and other observers have noted that SOPAMIB has released few details of Wahgnion's and Boungou's operational and financial performance since restart; most of what is public comes from official announcements.

This is the standard state-enterprise problem and it has a standard remedy: an external auditor with its own data. The Cour des comptes is the constitutional body for that role. What it needs is not more authority but its own feeds: the refinery's receipts of SOPAMIB doré, SONABEL's power records for the mines, customs entries for cyanide and other reagents, and satellite imagery of the pits, none of which pass through SOPAMIB. With those, the physical-balance reconstruction of section 3.3 can be run on the state's own mines by a body outside the ministry, and the Treasury can take SOPAMIB's production figures as verified rather than as reported. Section 5 places the audit engine under the Cour, and Part II specifies the feeds.

2.5 Investor exit and the exploration pipeline

Nationalisation changes who owns today's mines. It does nothing for tomorrow's unless someone keeps exploring. Burkina Faso's industrial output was flat at about 41 tonnes in the first nine months of 2025, after three years of decline, and the growth in 2025 came almost entirely from the artisanal sector. IAMGOLD's US$165 million of 2026 capital spending at Essakane is maintenance to hold production at 400,000 to 440,000 ounces, not expansion. Orezone's new plant at Bomboré and West African Resources' Kiaka are the only new industrial capacity, and both were financed before the 2024 code. Juniors, who do most greenfield exploration in West Africa, have largely stopped drilling in Burkina Faso, partly because of security and partly because a discovery may be nationalised.

The three exploration licences that SOPAMIB acquired with Wahgnion and Boungou illustrate the problem. An exploration licence is a right to spend money looking for gold; it is worth exactly what is spent on it, discounted by the chance of finding something. Unless SOPAMIB has an exploration budget, geologists and drilling contractors, the licences are worth nothing, and the state will find in ten years that it has nationalised a shrinking asset base. Section 8.7 proposes a valuation of the licences against a ring-fenced exploration budget with milestone reporting.

2.6 Artisanal gold

Burkina Faso's artisanal sector is large, old and, until 2024, almost entirely informal. Estimates of output ranged from twenty to forty tonnes a year, sold through Ouagadougou traders and across the Togo, Ghana and Niger borders to Dubai. Much of it comes from the north and east, where armed groups control or contest territory and where gold finances them. Recorded artisanal production in 2024 was 8.1 tonnes.

In 2025 it was about 42 tonnes. The change came from the creation of SONASP, the national precious-substances company, which set up buying counters across the country, paid promptly at close to world prices, and supported the formalisation of artisanal cooperatives into official marketing channels. The government attributes more than CFA 100 billion of additional revenue to the anti-fraud and formalisation effort. This is the largest single-year formalisation of artisanal gold on record anywhere in the region, and it happened in a country with worse security than Ghana or Tanzania.

Two lessons follow for the paper. First, nationalisation of industrial mines and formalisation of artisanal purchases are separate policies with separate revenue consequences, and in 2025 the second produced more new tonnes than the first. Second, the counter terms did the work: SONASP's price and payment speed moved gold from smugglers to the state without a levy. The channel-choice model of section 3.5 explains why, and Part II uses the 2024 to 2025 jump to show what the model's parameters must be. The policy risk now is that the state, seeing 42 tonnes at its counters, taxes them, and the gold goes back where it came from. Section 6 and Part II argue for raising revenue from the sector through the counter margin and income tax on resolved operators rather than through a levy that reverses the formalisation.

Models

3.1 Model A: ownership versus taxation

The first model puts the ownership decision in a form that Parliament can examine. It compares two present values from the state's point of view. The first is what the state would receive from taxing a private operator: the fiscal take share g applied to the operator's cash flow, discounted at a rate appropriate to a stream of tax receipts. The second is what the state receives as owner: the dividends it declares to itself, net of the price it paid and the cost of restarting the mine, discounted at the state's own cost of capital. The comparison turns on a small number of parameters, and each is worth naming. Compare the state's net present value from a mine under private operation with fiscal take share g, and under state operation with cost drift lambda (state cost as a multiple of private cost), acquisition price B, and dividend payout ratio pi after reinvestment:

NPV^{tax} = \sum_t \frac{g\, \Pi^{priv}_t}{(1+\delta)^t}, \qquad NPV^{own} = -B + \sum_t \frac{\pi \left( p_t q_t - \lambda\, c_t q_t - K_t \right)}{(1+\delta_s)^t}

Cost drift lambda is the ratio of the state operator's cost per ounce to what the private operator achieved on the same ore: 1.0 means SOPAMIB runs the mine exactly as Endeavour did, 1.1 means 10% more expensive. Payout pi is the share of after-cost cash the state actually takes as dividend rather than leaving in the company; a state company that retains cash for expansion, for arrears to suppliers or for reasons of its own pays the Treasury less than the mine earns. The state's discount rate delta-s exceeds the private one when the sovereign is cut off from concessional finance, which raises the bar for ownership: a dollar of dividend in year five is worth less to a state borrowing at 12% than the same dollar of tax would have been worth to a state borrowing at 8%. The break-even cost drift lambda-star at which the two are equal is the key number for every future nationalisation decision:

\lambda^{*} = \frac{p q - K - \frac{1}{\pi}\left( g\,\Pi^{priv} \frac{(1+\delta_s)}{(1+\delta)} + B\, a \right)}{c q}

where a is the annuity factor converting the acquisition price B to an annual charge over the mine's life. The formula says that ownership breaks even when the margin the state keeps as owner, after paying for the mine, equals the margin it would have taken as tax collector, adjusted for the difference in discount rates. Every term in it is observable or controllable: price and cost from the technical report, the private take from the fiscal regime, the acquisition price from the decree, the payout ratio from a dividend policy the state can write, and the discount rate from the state's cost of finance. Part II runs the formula on Wahgnion's actual terms and finds break-even cost drift between about 1.01 and 1.36 at a 12% sovereign discount rate depending on price, and much higher with cheaper finance and full payout. The finding is that the state's own financial policy, its cost of capital and dividend rule, matters as much to the outcome as SOPAMIB's operating performance.

3.2 Model B: SOPAMIB fiscal transparency model

The second model is a reporting discipline rather than a calculation. It asks SOPAMIB to present its accounts each quarter as if it were still a private taxpayer, computing the royalty, corporate tax and minority dividend it would have paid under the previous regime, alongside the dividend it actually declares. The difference between the two totals, delta, is the annual answer to the question in section 2.1: is the state better off owning this mine than it was taxing it? Treat SOPAMIB as if it were a private taxpayer for reporting purposes. Compute notional royalty, notional corporate tax and the dividend separately, so that the Treasury can see whether the mine would have paid more under the old regime:

R^{notional}_t = r\, p_t q_t + t\, \Pi_t, \qquad D_t = \pi \left(1 - t\right) \Pi_t, \qquad \Delta_t = \left(R^{notional}_t + D_t\right) - R^{priv}_t

A negative delta for two consecutive years is the trigger for an external review of SOPAMIB's cost base. The reporting has a second use: Burkina Faso is an EITI implementing country, and the notional accounts fit the EITI reconciliation template, so they create an external check through an existing mechanism without new law. Part II works the notional accounts for a year at US$4,000, where ownership wins by a wide margin, and at US$2,500, where the gain falls to near zero before the acquisition charge and below it after; the point is that the comparison is produced every year, by SOPAMIB, in a form the Cour des comptes and Parliament can read, rather than argued once at the time of the decree.

3.3 Model C: physical-balance audit of the state operator

The third model is the physical-balance reconstruction of the series overview, applied by the Cour des comptes to the state's own operator. The method is the same as for a private mine: reconstruct ounces produced from cyanide, power, refinery receipts and satellite volumetrics, and score the declaration on how far below those estimates it sits. What differs is the requirement for independence. Because both the operator and the auditor are public bodies, the custodian records must come from third parties and must reach the Cour directly: the refinery that receives SOPAMIB's doré, SONABEL for power, customs for reagent imports, and the satellite operator. A feed that passes through SOPAMIB or the mining ministry is not independent, and an audit built on it will not be believed. The same indirect-method reconstruction used elsewhere in the series, applied by the Cour des comptes to SOPAMIB:

A_i = \sum_m \omega_m \max\!\left(0, \frac{\ln \hat{Q}^{(m)}_i - \ln Q^{dec}_i}{\sigma_{m,s}}\right) + \delta H_i

3.4 Model D: price-risk and budget exposure

The fourth model treats the state's gold income as what it now is: a portfolio of equity positions in mines, exposed to price and to security events. Under taxation the state's royalty income fell in proportion to price; under ownership its dividend income is a leveraged claim, because costs are fixed and the dividend is the margin. The dividend from a mine with cost c at price p is proportional to (p minus lambda c) when positive and zero otherwise, which is the payoff of a call option on the gold price with strike at the mine's cost. A budget that counts on that dividend at US$4,000 is short an option it does not know it has sold. The budget's exposure to gold is now direct. Model the dividend stream as an option-like payoff and size a stabilisation buffer from its distribution:

D_t = \pi \max\!\left(0, (p_t - \lambda c) q - K\right), \qquad B^{*} = \bar{D} - VaR_{0.05}(D)

The buffer covers the gap between the budgeted dividend and its bad-year value at the 5th percentile, and is funded from above-forecast dividends in good years. With three SOPAMIB mines and minority stakes in others, the distribution should be simulated on the portfolio, including the correlation of all mines with the gold price and a security-closure scenario in which one mine produces nothing for a year. The IMF programme's deficit ceiling gives the buffer a concrete purpose: it is the instrument that keeps the deficit within the ceiling when gold falls, and Part II sizes it for a US$1,000 price decline.

3.5 Model E: artisanal channel

The fifth model is the channel-choice model of the Ghana paper and the series overview, and it needs less argument in Burkina Faso than anywhere else because the country has just run the experiment. SONASP's counters raised formal artisanal purchases from 8 to about 42 tonnes in a year by changing the terms sellers face: price close to world, prompt payment, and a legitimate channel for cooperatives. In the model that is a large shift in the effective discount Delta and the delay d, and Part II shows what the shift implies for the sensitivity parameters b and c. Formal share is logistic in levy, price premium, payment delay and enforcement, and the levy is set to maximise revenue subject to a formal-share floor. In Burkina Faso the enforcement term is weak in insecure zones, so the price and payment terms carry the design, and the same parameters that explain the jump imply that a levy of more than about 1% would reverse a large part of it. The revenue case for the sector rests instead on the counter margin, on income tax from the operators that entity resolution identifies, and on provenance rules that keep gold from contested areas out of the counters without driving it back to the smugglers.

Algorithms and illustrative results

4.1 Algorithm: nationalisation decision

FOR each candidate mine:
  estimate private fiscal take g, NPV_tax across price distribution
  estimate acquisition price B, state discount rate delta_s, payout ratio pi
  FOR lambda in 0.95..1.30: compute NPV_own
  lambda_star <- cost drift at which NPV_own = NPV_tax
  IF SOPAMIB's demonstrated cost ratio at existing mines > lambda_star: do not acquire
  ELSE acquire with a published dividend policy and an external audit covenant

4.2 Illustrative results

Representative mine: 250,000 ounces a year, all-in cost US$1,600, price US$4,000, acquisition at 4 times annual free cash flow, payout ratio 60%, private take share 55%, private discount rate 8%, state discount rate 12%. Under these assumptions the break-even cost drift is about 1.10: if SOPAMIB runs the mine at more than 10% above the private operator's cost, the Treasury receives less than it would have from taxing the private owner. At a US$2,500 price the break-even falls to about 1.02, because thinner margins leave no room for cost drift. Each point of cost drift above break-even costs the state roughly 1.3 points of take. Figures are illustrative and depend on the acquisition price above all.

4.3 Reading the results

The single number in 4.2, a break-even cost drift of about 1.10 at US$4,000, says that state ownership of a representative mine earns the Treasury more than taxing a private owner only if the state operator's cost per ounce stays within 10% of the private operator's. Part II replaces the single number with a table on Wahgnion's actual acquisition terms, and the table changes the reading in two ways.

First, the break-even depends heavily on the gold price. At US$2,500 the room for cost drift nearly vanishes, at 1% to 7% depending on financing terms; at US$5,000 it widens to 36% or more. Ownership is a leveraged bet on the price, and the state is taking it at a moment when the price is near an all-time high. That is not an argument against the policy, but it is an argument for the buffer in section 3.4 and for the acquisition test in section 3.1 being run at a price well below spot.

Second, the state's own financial terms matter as much as SOPAMIB's operating performance. At a 12% sovereign discount rate and an 80% payout, the room at US$4,000 is 22%; at 10% and full payout it is above 60%. The difference is not the mine; it is the cost at which the state finances the acquisition and the share of the mine's cash it actually takes as dividend rather than leaving in the company. Both are policy choices. A published dividend rule and re-entry to concessional finance would each do more for the fiscal return on nationalisation than any improvement in mine management could.

The cost of getting it wrong is asymmetric. Each point of cost drift above break-even costs the state more than a point of take, because under ownership the state bears the full cost increase and receives only the dividend, whereas under taxation it shared the cost with the operator through the corporate tax base. A state operator that is 20% more expensive than the private one at a mine where break-even is 10% has not merely lost the difference; it has converted a tax-paying asset into a subsidised one.

4.4 What the illustrative numbers do not show

SOPAMIB's actual operating cost at Wahgnion and Boungou is not public, and neither is its dividend policy or the terms on which the acquisition was financed. Part II's table is therefore a set of conditional results: if the cost ratio is X and the financing terms are Y, ownership earns more or less than taxation. The first thing SOPAMIB could do to settle the question is publish the notional accounts of section 3.2 for 2025 and 2026, from which the cost ratio can be read directly.

The representative-mine calculation also treats the private alternative as available. Endeavour sold because it wanted to exit; the counterfactual private owner was Lilium, a Burkinabè company, whose ability to operate the mines was itself uncertain. The comparison the state faced in 2025 may have been between SOPAMIB and a mine that closed, not between SOPAMIB and Endeavour. In that case ownership was the only way to keep the mine producing, and the relevant question is not whether it beats taxation but whether the dividend covers the acquisition and restart. Section 8.2 proposes the counterfactual analysis explicitly.

Solutions, software, data science and AI models

Software stack. FMIS and mining cadastre as sources; Python with Airflow; PostgreSQL bitemporal warehouse; NumPy, SciPy, PyMC; scikit-learn, PyOD; Splink; Sentinel-2 pipeline with rasterio; Superset; Docker and MLflow; French-language interfaces; hosted in Ouagadougou with an encrypted replica abroad.

Governance of AI. The audit engine runs under the Cour des comptes, not the mining ministry or SOPAMIB. Its findings are published in summary. Acquisition appraisals go to Parliament before decree.

Team. A five-person mining fiscal unit in Finance and a three-person audit cell in the Cour des comptes, with 18 months of external support and a partnership with a peer supreme audit institution for training.

5.7 The solution described as a system

The system's organising principle is separation: the body that operates the mines, the body that reports on them, and the body that audits them must be different, and the audit body must have its own data. Everything else follows from that.

SOPAMIB reports. The notional-accounts module in the government financial management system computes, each quarter and for each SOPAMIB mine, the royalty and corporate tax the mine would have paid under the previous regime, the dividend it would have paid to a 15% state shareholder, the dividend it actually declares to the state as full owner, its capital calls, its supplier arrears and its tax arrears. It presents the total under ownership beside the total under the previous regime and computes delta. The module also records SOPAMIB's cost per ounce against the previous operator's published all-in sustaining cost, inflation-adjusted, which is the cost ratio the acquisition test needs. The output goes on the EITI reconciliation template, so that it is checked by an existing process.

The Cour des comptes audits. The physical-balance engine runs under the Cour, on feeds that reach it directly: SONABEL's power records for each mine; customs entries for cyanide and other reagents, linked to mine deliveries through the importers' sales records; refinery receipts of SOPAMIB doré from SONASP's refinery and any export refiner; and Sentinel-2 volumetrics for the open pits. The engine reconstructs ounces produced from each source, scores the declaration, and flags identity breaches. Its findings go to the Cour's report, to Parliament and, in summary, to the public. The engine is the same software as the Ghana and Mali papers'; what is distinctive is where it sits.

The Ministry of Finance decides. The acquisition appraisal tool runs the ownership-versus-taxation model of section 3.1 on any candidate mine, from its technical report and the offered price, at the state's cost of finance and dividend policy, across a price distribution, and reports the break-even cost drift at the 25th, 50th and 75th percentile prices. The Ministry compares the result with SOPAMIB's demonstrated cost ratio from the reporting module. The appraisal goes to Parliament before the decree. The buffer calculator sizes the stabilisation buffer from the portfolio of state gold cash flows and the security-closure scenario, and reports the buffer's balance against target each quarter.

SONASP buys and the counters feed the model. SONASP's purchase system records every artisanal transaction (counter, seller, weight, assay, price, payment time, provenance declaration). The channel-choice model is estimated on that data monthly, and the levy, if any, and the counter margin are set from it with a formal-share floor. The trade-mirror dashboard compares SONASP purchases and declared exports with Togo, UAE and Swiss import data to bound what still escapes and to check the provenance rules.

Two things the system deliberately does not do. It does not put the audit engine in the mining ministry or in SOPAMIB, because an audit run by the operator's supervisor is not an audit. And it does not automate any assessment: the engine ranks and flags, the Cour reports, and the Ministry and Parliament decide.

5.8 Cost, staffing and sequencing of the build

Two small units. A mining fiscal unit of five in the Ministry of Finance: a head, a data engineer, two analysts with mining-finance backgrounds, and a liaison to SOPAMIB's board. An audit cell of three in the Cour des comptes: an auditor with mining experience, a data engineer and a geospatial analyst, supported by a twinning arrangement with a peer supreme audit institution that has audited state mining companies. External support of about 18 months for the build and for training. Open-source software throughout; the cost is people, the twinning and satellite processing, and it is small against mining revenue of CFA 776 billion.

Sequencing: the notional-accounts module first, because it needs no new data, only SOPAMIB's own; the physical-balance engine's production module next, because SONABEL and customs data are obtainable; the acquisition appraisal tool early, because further nationalisations are announced and each should be tested before decree; the buffer once the first year of dividend data exists; the SONASP estimation whenever its purchase data can be extracted, which should be immediate.

Reform sequence

6.1 The phases explained

Phase 1, notional accounts (2026 Q4). SOPAMIB publishes, for Wahgnion and Boungou, the notional royalty, tax and minority dividend under the previous regime, the actual dividend, capital calls and arrears, and its cost per ounce against Endeavour's last published all-in sustaining cost, for 2025 and the first three quarters of 2026. This answers the question in section 2.1 with SOPAMIB's own numbers and needs no new law. Deliverable: the published accounts on the EITI template. Owner: SOPAMIB, under instruction from the Ministry of Finance. Risk: SOPAMIB reports late or partially; the EITI process and Parliament's budget committee are the pressure.

Phase 2, dividend policy and audit covenant (2027). A decree sets SOPAMIB's dividend policy as a formula (a fixed share of free cash flow after a defined sustaining-capital allowance) and requires an annual external physical-balance audit by the Cour des comptes with direct data feeds. The first audit runs on the 2026 declarations. Deliverable: the decree, the data-feed agreements with SONABEL, customs and the refineries, and the first audit report. Owner: the Ministry of Finance and the Cour. Risk: the feeds are routed through SOPAMIB or the mining ministry, which defeats the purpose; the decree must specify direct delivery.

Phase 3, the acquisition test (2027). The next nationalisation candidates are run through the appraisal tool at the state's actual cost of finance and the Phase 2 dividend policy, against SOPAMIB's demonstrated cost ratio from Phase 1, and the results are published and laid before Parliament before any decree. Deliverable: published appraisals. Owner: the Ministry of Finance. Risk: the test is run after the political decision rather than before; publication is the safeguard.

Phase 4, the stabilisation buffer (2027 to 2028). The buffer is legislated, its target set from the portfolio simulation with a security-closure scenario, and funded from dividends above forecast. Its balance is reported quarterly against the IMF programme's deficit ceiling. Deliverable: the buffer law and the first quarterly report. Owner: the Ministry of Finance. Risk: above-forecast dividends are spent rather than saved in a high-price year; the law's contribution rule is the constraint.

Phase 5, artisanal counters and mirror agreements (2028). SONASP's counter terms are set from the estimated channel-choice model with a formal-share floor; any levy is set well below the rate that would reverse the 2025 formalisation, and the revenue case for the sector shifts to the counter margin and to income tax on operators resolved from the purchase data. Mirror-data agreements with Togo and the UAE measure the leakage. Deliverable: the estimated model, the published counter rule, and the first leakage estimate. Owner: SONASP and the tax authority. Risk: a levy is imposed before the estimation, on the visible 42 tonnes; the model's implied sensitivity, worked in Part II, is the argument against it.

6.2 What can go wrong in the sequence

The first risk is that Phase 1 never happens, because SOPAMIB's performance is politically sensitive and the notional accounts would make it visible. Everything else depends on it, and the paper's design puts it first for that reason: it is the cheapest step and the one that reveals whether the rest is needed. The second risk is that the audit engine is built inside the mining ministry, which would produce a report no one believes. The third is that the acquisition test is used to justify decisions already taken, rather than to inform them; publishing the appraisal before the decree is the only protection. The fourth is a price fall before the buffer exists, which would convert the state's gold exposure into a budget shock with nothing set aside; the buffer should therefore be started from the first above-forecast quarter, before its law is complete.

Limitations

7.1 The limitations discussed

Unpublished cost, price and policy. The paper's central calculation, the break-even cost drift, needs SOPAMIB's operating cost, the financing terms of the acquisition and the dividend policy. None is public. Part II runs the calculation on the public acquisition price and a range of assumptions for the rest; its result is a table of conditions, not a verdict. The paper's first recommendation, the notional accounts, exists to supply the missing inputs.

Security. Some mines may be unoperable under any owner for periods the model cannot predict. The security-closure scenario in the buffer model is a crude representation of that risk, calibrated on Boungou's 2019 closure; a longer or more widespread disruption would exceed any buffer sized this way. The model can price a known risk; it cannot price the absence of the state.

The sovereign discount rate. Burkina Faso's cost of finance while outside concessional and regional markets is uncertain and probably higher than the 12% the paper uses as its upper case. A higher rate lowers the break-even further. Re-entry to concessional finance would change the fiscal case for ownership more than any operational improvement, and it depends on political decisions outside the paper's scope.

Objectives beyond revenue. Nationalisation is framed by the government as sovereignty, local employment and control over a strategic resource. The paper weighs none of these; it asks only what ownership does to the Treasury. A government that values the other objectives may accept a fiscal cost for them, and the paper's contribution is to make that cost visible so that the choice is made knowingly.

The counterfactual owner. The comparison assumes a private operator was available to tax. At Wahgnion and Boungou the seller wanted out and the buyer was untested; the real alternative may have been closure. Section 8.2 proposes estimating the counterfactual explicitly, using Endeavour's sale process as the market's valuation.

Artisanal output. The 42 tonnes purchased by SONASP in 2025 is a purchase figure, not a production figure. How much of it was newly formalised Burkinabè production, how much was previously smuggled output redirected, and how much originated across a border or in contested areas is unknown. The mirror and footprint analysis in section 8.4 is proposed to find out; until it does, the formalisation's revenue value is a range.

Extended modelling: improvements and further research

8.1 Research update, September 2026: the numbers have changed

The 57-tonne figure in the abstract is out of date. The Mines Minister reported national gold production above 94 tonnes in 2025, up from about 61 to 63 tonnes in 2024, with roughly 42 tonnes from artisanal and semi-mechanised mining, against 8.1 tonnes in 2024. The jump is almost entirely formalisation: SONASP, the state precious-substances company, set up buying counters and now purchases artisanal output, and the government reports more than CFA 100 billion in additional revenue from anti-fraud efforts. Mining budget revenue reached CFA 776 billion in 2025. Industrial output was flat at about 41 tonnes for the first nine months. The IMF notes the fiscal deficit stayed within the 4% of GDP programme objective in 2025 largely on gold revenue. Three mines are now under SOPAMIB and six of fifteen industrial mines are majority Burkinabè-owned; the state holds 15% of Essakane, where IAMGOLD committed US$165 million of 2026 capital spending. The Wahgnion and Boungou acquisition terms are public: US$60 million to Endeavour plus a 3% royalty on up to 400,000 ounces sold from Wahgnion. SOPAMIB has released few operational or financial details on either mine. For the paper: the artisanal channel model is now the largest revenue story, and the acquisition price for the break-even calculation is known.

8.2 Model A, ownership versus taxation: improvements

8.3 Model B, SOPAMIB transparency: improvements

8.4 Model C, physical-balance audit: improvements

8.5 Model D, budget exposure: improvements

8.6 Model E, artisanal channel: improvements

8.7 New models proposed

8.8 Further research

Sources for this section: APA News, Burkina: record 94 tonnes in 2025; Ecofin Agency, artisanal gold's share of output, 5 February 2026; Ecofin Agency, production up 58% in nine months, December 2025; Ecofin Agency, state-owned mines restart, transparency questions; Invest-Time, gold and the Burkinabè economy; The Exchange Africa, Burkina Faso mining 2026.

A. Worked example 1: Wahgnion on the actual acquisition terms

The acquisition price is public: US$60 million to Endeavour plus a 3% royalty on up to 400,000 ounces sold from Wahgnion, which at US$4,000 is worth about US$48 million. Assume a restart cost of US$40 million, 150,000 ounces a year for eight years, all-in sustaining cost US$1,350 under the previous operator, and a private fiscal take of 55%. The state's alternative is to have taxed a private owner (NPV at 8%) or to own the mine (NPV at the sovereign's higher discount rate, with a dividend payout ratio pi). The break-even cost drift lambda-star is the multiple of private cost at which the two are equal.

Reading: at a 12% sovereign discount rate and 80% payout, ownership breaks even only if SOPAMIB runs Wahgnion within 1% of Endeavour's cost at US$2,500, within 8% at US$3,000 and within 22% at US$4,000. Cheaper sovereign finance and a full payout widen the room considerably. The two variables the state controls, its cost of capital and its dividend policy, matter as much as the mine's operating performance. The 10% figure quoted in the abstract sits inside this range at the price and terms assumed there; the table replaces it.

B. Worked example 2: SOPAMIB notional accounts

For a year in which Wahgnion produces 150,000 ounces at US$4,000 and US$1,500 all-in cost under state operation (an 11% cost drift), the notional accounts under the previous private regime (5% royalty, 27.5% corporate rate, 15% state equity, figures illustrative) and the actual state result are:

Ownership wins comfortably in this year at this price. The same table at US$2,500 gives revenue of 375, state-operation profit of 150 and a dividend of about 87, against a private-case total to the state of about 62 plus royalty 19 and tax 42, and the net gain falls to roughly 20 before acquisition and turns negative after it. The point of the notional accounts is that this comparison is produced every year, by SOPAMIB, in a format the Cour des comptes and Parliament can read.

C. Worked example 3: the artisanal formalisation jump

SONASP purchases rose from 8.1 tonnes in 2024 to about 42 tonnes in 2025. Treating total artisanal output as roughly stable at 45 to 50 tonnes (a figure that needs the mirror and footprint estimate of the overview, section 3.2), formal share rose from about 0.17 to about 0.88 in one year. In the channel-choice model that is a shift in the log-odds of about 3.6. If the price gap moved from 6% to 1% and payment delay from 10 days to 1 day, the implied coefficients are of the order b = 0.4 per point and c = 0.2 per day, larger than the Ghana defaults; if enforcement (rho) also rose, they are smaller. The 2025 data therefore identifies the model well, and the paper's recommendation for the levy follows: with b around 0.4, a levy of 1.5% costs about 0.6 log-odds, or roughly 12 points of formal share from 0.88, which is a large loss; a levy of 0.5 to 1% costs 4 to 8 points. Burkina Faso's revenue from artisanal gold is better raised through the counter margin and the income tax on resolved operators than through a levy that reverses the formalisation.

D. Algorithms in full

D.1 Acquisition test

ALGORITHM Acquisition_Test
INPUT: candidate mine technical report; offered price B; restart cost; private regime terms; sovereign discount rate; payout policy
STEP 1  NPV_tax across price distribution under private ownership and current regime
STEP 2  For lambda in 0.9 .. 1.6: NPV_own(lambda) across the same price paths, net of B, restart and security cost
STEP 3  lambda_star(price percentile) where NPV_own = NPV_tax; report at 25th, 50th, 75th percentile prices
STEP 4  Compare lambda_star with SOPAMIB's demonstrated cost ratio at operating mines (from D.2)
STEP 5  Decision rule: acquire only if demonstrated ratio < lambda_star at the 25th percentile price
STEP 6  Publish the test result before decree

D.2 Notional-tax reporting

ALGORITHM Notional_Accounts
QUARTERLY for each SOPAMIB mine:
  compute revenue, cost by category, notional royalty, notional CIT, notional dividend at the previous private equity share
  compute actual dividend declared and paid; capital calls; supplier arrears; tax arrears
  cost ratio = state cost per ounce / previous operator's published AISC (inflation-adjusted)
  delta = (notional royalty + notional CIT + actual dividend) - private-case total to state
PUBLISH on EITI template; two consecutive negative deltas trigger external cost review

D.3 External physical-balance audit

ALGORITHM External_Audit (run by the Cour des comptes)
INPUT: SOPAMIB declared tonnes, grade, recovery, ounces; SONABEL power; customs reagent imports; refinery receipts; Sentinel-2 volumetrics
FOR each mine: Q_hat by source; z-scores; hard flags on refinery > declared and throughput > power-implied
REPORT: reconstruction band per mine; any breach; recommended adjustments to notional accounts
INDEPENDENCE: data feeds go directly to the Cour, not via SOPAMIB or the mining ministry

D.4 Budget buffer

ALGORITHM Gold_Buffer
INPUT: portfolio of state gold cash flows (SOPAMIB dividends, minority dividends, royalties, SONASP margin); price distribution; security-shock scenario
SIMULATE annual state gold cash under 10,000 price paths and a 10% annual probability that one SOPAMIB mine is closed for a year
B_star = budgeted gold cash - 5th percentile of simulated gold cash
CONTRIBUTION: above-forecast gold cash goes to the buffer until B_star is reached
DRAWDOWN: only when gold cash falls below budget; report to Parliament

E. Data specification for Burkina Faso

F. Sensitivity analysis

G. Validation protocol

References & keywords

• EITI — Burkina Faso country profile ↗
  State participation and extractive-sector transparency
• EITI — Burkina Faso 2024 report ↗
  SOPAMIB status and mining-code context
• West African Development Bank — Boungou and Wahgnion ↗
  2025 approved financing for mine operating capacity
• Endeavour Mining — 2023 asset sale ↗
  Original Boungou and Wahgnion transaction context
• World Bank — Corporate Governance of State-Owned Enterprises ↗
  2014 toolkit on ownership and accountability
• IMF — Fiscal Regimes for Extractive Industries ↗
  2012 policy paper on mining fiscal design