We read last week that the Brazilian economy ministry had adjusted its monthly export data for September, October and November twice in under a week. The revisions were sizeable: that for November was close to US$6bn. They undermine the credibility of the statistical authority but also carry the risk of contagion. The GDP report for the third quarter was released after the first but before the second change to the export figures, so it may also have to be revised to allow for the new numbers for September.

 

There are repercussions for financial markets. Investors would have taken positions on the exchange rate on the basis of subsequently revised export figures. We have broader concerns about the release of data very quickly after the period that they cover. This is common practice, for example, in China, where several reputable analysts are uneasy with the national accounts. Markets and the public would be happy to wait a few more days for the data release and feel more comfortable with statistics.

 

Nigeria’s President Muhammadu Buhari appeared to be aligned with the right side when he told his new Economic Advisory Council on 09 October to focus on primary data collection. He then went off-script by calling for Nigerians to rely on domestic data sources rather than the series from the IMF, World Bank and other international bodies.

 

This intervention overlooked the reality that the macroeconomic data of the Bretton Woods duo is drawn from national sources. The Bank and the Fund can advise on methodology and provide technical support but do not have the remit or the resources in the field to produce their own series for GDP or inflation. Buhari did not seem aware that the National Bureau of Statistics, which reports directly to the presidency, has been receiving its government funding in arrears and therefore been unable to produce some of its regular reports (on unemployment, among others).

 

The quality of statistical provision varies across the continent. The best is South Africa, which reflects the fixation of the previous minority regime with keeping records. Also, the country’s financial markets are easily the most developed in Africa, and demand a high quality of data.

 

Elsewhere the quality is undermined by large informal economies and the prevalence of payment in cash. Financial inclusion in Nigeria has risen from 45.4 per cent in 2016 to 63.6 per cent in 2018. Its government has a target of 95 per cent for 2024, and knows well that the process creates new taxpayers as well as opportunities for low-income citizens such as buying livestock, acquiring land and building a house.

 

Technology plays a major role in lifting the inclusion rates and thereby facilitating statistical collection. It has also spawned a new genre of data collection firms, which combine traditional methods with geospatial sources and tech to track consumer spending. One such is the US-based Fraym, which produced a report last year (Finding the Dynamic African Consumer) to quantify premium (and other) consumers across the continent. Its findings are easily monetized, being of obvious interest to FMCG operators.

 

Another firm is Reach Technologies, which drew upon eight million data points from users of its app to chart the top ten transactions across more than 50 Kenyan and Nigerian cities in 2018.

 

In the name of even-handedness, we should note that developed economies also suffer statistical embarrassments. It emerged in Japan in January that national wage data had been understated for 15 years, and that an estimated 20 million people had not made adequate payments for unemployment insurance.

 

The world of statistics is constantly evolving. A newish trend is the gathering of “big data” so as to understand our expanding services sector economies. A good example would be the collection of VAT returns.

 

While the quality is improving, so does the horizon remain far in the distance. In the ideal world, we would add together all the balance-of-payments surpluses and deficits across the world, and arrive at a figure of zero. Sadly we don’t, and are confronted with large figures for “net errors and omission’’ (balancing items, broadly).

 

Jean Puri