How Not to Backcast Time Series Data, Or Why Britain’s Post-War National Accounts Could Still Lead You Astray

Eleven years ago, the Office for National Statistics, Britain's main statistics agency, attempted to merge 2 sets of time series data in order to backcast a long history of the country's capital investment. The restored investment figures became the series for total investment in the official 'historic' national accounts between 1948 and 1996. The ONS chose to merge old investment data with new national accounts data using splicing, a common technique, and did so 'bottom up': by adding up detailed backcast figures to derive the total. This paper, a sequel to one published in September 2024, argues that the ONS methodology was a mistake. Extraordinary differences between some of the old and new data series should have alerted the ONS to the deficiency of its approach. And the agency failed comprehensively to sense check its results. The implausible rewrite of Britain’s economic past that resulted is still embedded in the national accounts, but might be corrected as a result of a new ONS investigation begun in response to the earlier paper. Official correction is not assured, however. As matters stand, those wishing to draw lessons from Britain’s economic past are still advised not to rely on the 'historic' national accounts.