Open data: UK macroeconomic series from 1086 to present. Bank of England Millennium of Macroeconomic Data (Quarter)

Rows: 19,655
Columns: 8

Hundreds of annual time series covering nominal and real GDP, CPI, wages, interest rates, public debt, money supply, and equity markets for the UK. Some series go back to 1086. Most are continuous from at least 1700.

AI analysis

Dataset Analysis: Bank of England Millennium of Macroeconomic Data

1. Dataset Overview and Research Value

This dataset represents one of the most historically comprehensive macroeconomic archives publicly available, spanning UK economic indicators from 1695 to 2016 in quarterly format — with the broader series (described as reaching back to 1086) condensed into this quarterly structure. Published by the Bank of England as part of their "Millennium of Macroeconomic Data" project, it covers nominal and real GDP, CPI, wages, interest rates, public debt, money supply, and equity markets across 48 distinct variable IDs. No source URL was provided, but the Bank of England as publisher lends strong institutional credibility and methodological rigor to the series. The dataset's extraordinary temporal depth makes it exceptionally valuable for long-run correlation studies — for example, examining how interest rates relate to inflation cycles over centuries, or how money supply growth precedes debt accumulation across different political eras.

2. Data Quality Observations

Data quality is notably strong for a historical dataset of this scope. Total null cells across all 8 columns register at zero, which is remarkable given the 19,655-row count and the challenge of reconstructing pre-industrial economic data. No type mismatches are apparent from the column definitions — integers, strings, and decimals are each assigned to appropriate columns. Duplicate row counts are flagged as pending recomputation (Phase D), which is an important caveat: until duplicates are confirmed, row-level uniqueness cannot be guaranteed, and analysts should validate uniqueness on the composite key of (year, quarter, variableid) before aggregating. The period column having 1,288 distinct values against only 322 distinct years and 4 quarters suggests it may encode a formatted label (e.g., "1695 Q1") rather than a raw date — worth inspecting for parsing consistency across centuries.

3. Key Column Distributions

The value column is analytically the most complex and deserves close attention. Its range spans -16,750 to 2,246,470, with a mean of 37,370 wildly exceeding the median of 7.30 — a stark signal of extreme right-skew (skewness = 8.23). The interquartile range (Q1 = 3, Q3 = 1,796) confirms that the bulk of observations cluster near zero or small values, while 4,286 outliers (roughly 22% of rows) pull the distribution dramatically rightward. This is expected behavior in a long-run mixed-units dataset: nominal GDP in modern millions sits alongside interest rates expressed as percentages in single digits — all in the same value column. Analysts must filter by unit or variableid before any statistical work. The year column shows a mean of 1937.6 with a negative skew of -1.26 and 833 outliers, consistent with data density being much higher in recent centuries (quarterly resolution for recent years vs. sparser early coverage), and Q1/Q3 of 1903/1991 confirm most data mass falls in the 20th century.

4. Recommended Join Key Columns

The natural composite join key is year + quarter, which together define a unique time period. For cross-variable analysis within this dataset, variableid (48 distinct values) is more reliable than variable (46 distinct values) — the two-count discrepancy suggests possible label variants or aliases that could cause silent mismatches on string joins. For joining to external datasets, a standardized date column derived from year+quarter (e.g., "1695-Q1" or an ISO date like 1695-01-01) would be the most interoperable key. The section column (5 distinct values) serves well as a filter key to scope joins to relevant economic domains (e.g., joining only the monetary policy section to a central bank decisions dataset).

5. Recommended Pairing Datasets for Correlation Discovery

Several dataset categories would pair powerfully with this data:

- UK Political and Parliamentary records — correlating GDP growth, inflation, or public debt spikes with changes in government, wars, or major legislation (e.g., the Reform Acts, World Wars, Thatcher-era privatizations) - Global commodity price indices (oil, wheat, gold) — to study how external supply shocks transmit into UK CPI and wages over multi-decade horizons - Other central bank historical datasets (e.g., US Federal Reserve FRED long-run series, Sveriges Riksbank data) — enabling cross-country interest rate and inflation correlation analysis across similar time horizons - UK trade and balance of payments historical data — examining how money supply or exchange rates relate to trade balances across centuries - Epidemiological or climate datasets (e.g., harvest failure records, plague years, temperature reconstructions) — for novel long-run causal analysis linking environmental shocks to economic contractions, particularly relevant for the pre-1700 annual extensions of this series

Columns

year (integer)
quarter (integer)
period (string)
variable_id (string)
variable (string)
section (string)
unit (string)
value (decimal)