Methodology
The Aso Rock Index (ARI)
The Aso Rock Index is a composite measure that tracks how well Nigeria converts governance into economic outcomes. It combines four weighted components into a single annual score, with the year 2000 set to 100 as the baseline.
Formula
ARI = 0.40×G + 0.30×E + 0.15×D + 0.15×W
Components
G: Governance Quality (40%)
World Bank WGI composite score, normalized to 0-100 scale where 0 = worst possible (-2.5) and 100 = best possible (+2.5).
E: Economic Performance (30%)
Real GDP per capita (constant 2015 US$), indexed to 2000 = 100. Uses per-capita output so a larger-but-poorer economy does not score higher.
D: Diversification (15%)
Non-oil sectors (services, manufacturing, agriculture) as percentage of GDP. Higher = less oil-dependent.
W: Welfare Progress (15%)
Life expectancy at birth, indexed to 2000 = 100. A direct welfare measure.
Baseline and interpretation
- Baseline: Year 2000 = 100 (first full year of Fourth Republic)
- Above 100: Better governance-economic performance than baseline
- Below 100: Worse than baseline (military-era levels)
- All-time high: 136.8 (2015, end of the Jonathan era)
- All-time low: 99.8 (2002, Obasanjo)
Weight sensitivity: does the shape survive different weights?
The 40/30/15/15 weights are an editorial choice, and composite indices with hand-set weights are fairly criticised for it. So we tested whether the index's shape depends on the exact weights. We recomputed the ARI under twelve alternative weightings: each component's weight moved up and down by 10 points (with the others renormalised), plus equal weights (25/25/25/25), a governance-heavy split (55/25/10/10), and an economy-heavy split (25/45/15/15).
The result is reassuring for the index's central claims:
- The peak year is 2015 under every single weighting tested. Not most. All twelve, plus the baseline. The identification of 2015 as Nigeria's best combined governance-and-economy year does not depend on the weights.
- The trough is 2000 or 2002 under every weighting (2000 is the baseline year, fixed at 100 by construction; 2002 is the WGI's worst reading). No weighting produces a different trough.
- What the weights do change is the index level (the economy-heavy split lifts the 2015 peak; the governance-heavy split lowers it), not the ranking of years. The story the index tells is robust; the exact number on the y-axis is not load-bearing.
The sensitivity run reproduces the ranking from cached component data (governance, GDP per capita, sector shares) with life expectancy interpolated between its published World Bank anchor values. Index levels in the sensitivity run differ slightly from the committed vintage because of that reconstruction; the robustness of the peak and trough years is the finding, and it holds regardless. The full table is published as ari_sensitivity.json.
Caveats
Important limitations: The ARI reflects correlation, not causation. Oil price shocks affect all components simultaneously. Some years have missing governance data that is interpolated. The index is designed for tracking Nigeria's progress over time, not for cross-country comparison. The weights are editorial; see the sensitivity analysis above for what that does and does not change.
Data sources
Primary indicators come from official World Bank databases: a standardised methodology across countries and years, published openly, and refreshed on a regular cadence. The site is re-pulled quarterly and the data diary records every change.
World Development Indicators (WDI)
Economic data: GDP, per capita income, sector output, poverty rates
databank.worldbank.org/wdi →Worldwide Governance Indicators (WGI)
Governance scores: Government effectiveness, rule of law, corruption control
info.worldbank.org/governance/wgi →V-Dem (Historical Context)
Democracy indicators for pre-1996 period (military vs civilian rule classification)
v-dem.net →Full series registry
Every number on this site traces to one row of the registry below: the exact series, its code, source, unit, and vintage. "World Bank" is not a citation; "World Bank WDI, NY.GDP.PCAP.KD" is. Download the machine-readable version: sources.csv.
| Series | Code | Source | Unit | Vintage |
|---|
Historical context: pre-1996
World Bank governance data only begins in 1996. Nigeria's economic record runs much further back. For the pre-1996 baseline we draw on the Central Bank of Nigeria Statistical Bulletin series for 1990 through 1998, and on V-Dem democracy indicators for the military period.
The military era (1970-1999)
The pre-democracy period was marked by chronic price instability:
- Average inflation: 23% across military regimes
- Abacha era (1993-98): 39% average inflation, the worst of any sustained period
- 1995 peak: 73% inflation, during international sanctions
The SAP effect (1986)
The Structural Adjustment Program, imposed under IMF pressure during the Babangida regime, is central to any account of Nigeria's economic trajectory. The before/after on price stability and the naira:
Pre-SAP (1980-1985)
Average inflation: 17.8%
Naira roughly at parity with dollar
Post-SAP (1986-1993)
Average inflation: 27.1%
Naira crashed from ₦1=$1.50 to ₦4=$1
SAP eroded the middle class and bequeathed the structural conditions that democracy inherited in 1999. Obasanjo's recovery should be read against that low starting point.
Source: CBN Statistical Bulletin
Inflation data extracted from official Central Bank of Nigeria publications (1990-1998). These documents are available from the CBN Digital Commons archive.
dc.cbn.gov.ng/cbn_statistical_bulletin →Constant 2015 US dollars
GDP can be reported in two ways, and the choice matters:
Current/nominal USD
Moves with inflation and the exchange rate. Nigeria's GDP can rise on paper simply because the naira weakened or US prices rose. Misleading for cross-year comparisons.
Constant 2015 USD
Adjusted for inflation and locked to 2015 prices. A doubling reflects a doubling in real output, not in prices.
We use constant 2015 USD throughout, which is why our totals can differ from headline figures quoted in current USD. Both are valid; they measure different things.
How we calculate growth
GDP added
Simple subtraction: GDP at end of term minus GDP at start.
GDP Added = GDP(end year) − GDP(start year)
Obasanjo: $306.3B (2007) − $167.7B (1999) = $138.6B added
CAGR (compound annual growth rate)
Average yearly growth rate, accounting for compounding. This is the fairest way to compare presidents who served different lengths.
CAGR = (End Value / Start Value)^(1/years) − 1
Obasanjo: ($306.3B / $167.7B)^(1/8) − 1 = 7.7% per year
Per-capita change
GDP per person at end minus GDP per person at start. This accounts for population growth.
Per Capita Change = GDP/capita(end) − GDP/capita(start)
Buhari: $2,281 (2022) − $2,586 (2015) = -$305 per person
Statistical methods
Beyond the descriptive tables and charts, several statistical techniques are used to interrogate the governance-growth relationship. Each is applied selectively and only where the data and time horizon support it.
Time-series analysis
Stationarity, decomposition, and trend work uses standard statsmodels routines, plus Facebook Prophet for decomposition and trend extraction:
- Augmented Dickey-Fuller (ADF). Tests whether the GDP and governance series contain unit roots.
- KPSS. A complementary stationarity test used alongside ADF.
- Granger causality. Tests whether past governance values help predict future GDP, and the reverse.
- Seasonal decomposition. Separates trend, seasonal, and residual components.
- Facebook Prophet. Used for trend decomposition on the longer macro series.
Causal inference
Establishing causation requires more than correlation. The methods implemented in src/models/causal_inference.py and scripts/stage2_causal_inference.py are:
- Difference-in-Differences (DiD). Compares Nigeria to control countries around regime transitions and policy shifts to estimate treatment effects.
- Synthetic control. Constructs a "synthetic Nigeria" from weighted combinations of comparator countries to estimate counterfactual trajectories.
- Regression discontinuity. Exploits sharp policy or regime changes to estimate local causal effects.
Counterfactual and elasticity estimates
- Log-linear elasticity. Estimates how sensitive each sector is to changes in the governance score.
- Scenario projection. Applies estimated elasticities to alternative governance trajectories (e.g. "GDP path if WGI had held at the 2007 level").
- Bootstrap confidence intervals. Quantifies uncertainty around point estimates using 1,000 resamples.
Limitations
- All models assume approximately linear relationships; the underlying dynamics are not strictly linear.
- Nigeria's oil dependence limits the comparability of any control or benchmark set; synthetic-control and DiD weights should be read with that in mind.
- Roughly three decades of data is short for high-confidence causal inference. Estimates carry wide bands.
- The site does not currently implement instrumental-variable estimation, which would strengthen any causal reading of governance on growth.
Levels vs changes: why a correlation can lie
When two numbers both rise over time, they will correlate, whether or not one has anything to do with the other. Ice-cream sales and drownings both rise in summer. The honest test is not whether the levels move together, but whether the year-to-year changes do. We report both, and we lead with the de-trended figure.
The clearest case on this site is agriculture. We built a single series for total farm output (every priced commodity multiplied by its constant producer price) and set it against the governance composite, 1996 to 2024.
Farm output more than doubled across the period, climbing at the same steady pace through governance's worst year (−1.48 in 2002) and its best (−0.99 in 2015) alike. The +0.57 is simply two upward drifts overlapping. On the de-trended test, agriculture is effectively decoupled from governance. It tracks land under cultivation and rainfall, not the quality of institutions in Abuja. This is a deliberate contrast with the headline GDP-governance correlation (r≈0.68): the governance dividend, where it exists, sits in manufacturing, services and revenue management, not the farm. The full chart lives in the Commodities section of the dashboard.
Read this carefully
- The governance composite moved only about 0.5 points on a 5-point scale across the whole period. With so little variation in the input, this is best read as "no evidence of a link," not proof that none exists.
- "Decoupled" applies to aggregate output. It does not rule out governance effects on specific commodities, input markets, or rural security.
The counterfactual model
The governance-gap estimate referenced elsewhere on the site is computed two ways: a single-country regression (about $0.6T) and a cross-country panel (about $0.9T). The single-country version is a simple three-step calculation:
-
Estimate the relationship between governance and GDP using a cross-country regression.
Reading: every 1-point improvement in WGI governance is associated with roughly $592B more GDP in the comparison set.
-
Measure Nigeria's governance gap against a benchmark (Brazil).
Gap: Nigeria (-1.04) vs Brazil (0.00) = 1.04 points.
-
Multiply the gap by the coefficient to estimate the GDP shortfall.
1.04 governance points × ~$592B per point ≈ $0.6T (single-country). The cross-country panel gives ~$0.9T. Both are correlational with wide bands.
Read this as a scenario, not a finding
This is the single weakest-supported number on the site, and we would rather say so than have someone else say it for us. It is a scenario exercise, not a measurement.
- The 95% confidence interval includes zero. It runs from -$113B to +$1.2 trillion. A statistician reads that as: the estimate is not distinguishable from no effect at conventional levels. The point estimate is a midpoint of a very wide band, not a reliable figure.
- R-squared is 0.13. Governance explains about 13% of the variation in GDP over the period. The other 87% is everything else, oil first among them.
- This is a correlation, not a causal estimate. We do not claim governance alone produced the gap. Our own Granger tests (p-values of 0.28 to 0.51) cannot establish the direction of the relationship.
- Oil prices confound the relationship; high-growth years often coincided with high oil prices.
- The benchmark choice matters. Substituting Brazil for a different reference country produces a different number.
We keep the estimate on the site because "what might a peer-level governance path have been worth" is a question worth posing, and because the order of magnitude (hundreds of billions) is defensible even if the point estimate is not. But it is a supporting exhibit, not a headline. The site's load-bearing findings are the descriptive ones: the naira collapse, the essentials-versus-CPI gap, and the harvest-versus-food decoupling, none of which rest on a causal model.
Governance score
The World Bank's WGI scores run from -2.5 (worst) to +2.5 (best).
The composite averages four components:
- Government effectiveness. Capacity of the state to deliver public services.
- Regulatory quality. Whether the policy environment supports or impedes private activity.
- Rule of law. Enforceability of contracts and impartiality of the courts.
- Control of corruption. The extent to which public office is used for private gain.
Measurement error: read small changes with caution
The WGI is not a direct measurement. It is a weighted average of underlying expert and survey sources, and the World Bank publishes a standard error alongside every score for exactly this reason. For a country like Nigeria the standard error is typically in the range of 0.15 to 0.25 points per year.
That matters for how this site should be read. Several of the per-president governance changes we report are within the measurement noise of the instrument:
- Obasanjo +0.203 and Jonathan +0.208 are the only two moves large enough to be plausibly distinguishable from zero, and even those are marginal once the source uncertainty is stacked across a term.
- Yar'Adua -0.134, Buhari -0.077, and Tinubu -0.012 are all comfortably inside a single year's standard error. We do not treat them as statistically established declines.
The honest reading of the record is not "two presidents improved governance and three worsened it" as a precise ranking. It is that measured governance barely moved in twenty-eight years, and that only Obasanjo's and Jonathan's improvements rise even tentatively above the noise floor. Kaufmann and Kraay, who built the WGI, make the same warning: the indicators are designed for broad cross-country and long-horizon comparison, not for reading small year-to-year or term-to-term differences as signal.
Where per-president governance changes appear on this site, they should be read as indicative, not as measurements with the precision the decimal places imply.
Out of scope
- Causal claims. We surface patterns and associations, not proof that governance caused a given outcome.
- Oil-price adjustments. Any normalisation here would rest on assumptions we are not in a position to defend.
- Nigerian government statistics as primary source. Headline figures come from internationally-standardised World Bank series; CBN data is used for pre-1996 context.
- Forecasting. The site documents the historical record; it is not a forward projection.
- Apportioning blame. External factors (oil shocks, global crises, security shocks) sit alongside policy choices in every administration.
Reproducibility
All headline indicators on the site can be reproduced from the public sources listed above. Raw series pull directly from the World Bank DataBank (codes in each source card). Indicator definitions and any transformations are documented in this page and in the data diary. Queries, corrections, or requests to see analysis code: [email protected].