Last week I wrote a blog post titled “Africa is rising, employment is not” reporting some figures from a new paper by John C. Anyanwu  on the African Statistical Journal. I briefly mentioned in the post that employment stats are not very reliable in Africa because of informality of the labour market. But I kept it a bit vague and some readers put me on the spot: if the data is garbage, why are you blogging about it?

Good question. The reality is that I knew that the data was “not very reliable”, but I didn’t  look in-depth into the issue -so I’ll try to do this now. Look at these national-level definitions taken from the ILO laborista database:

In Kenya, employment data is taken from this question in the national census: 

What was X mainly doing during the last seven days preceding the census night? 1) worked for pay or profit; 2) on leave/sick leave; 3) working on family holding; 4) no work; 5) seeking work; 6) student; 7) retired; 8) disabled; 9) home makers; 10) other

A person is considered employed if over the last week she/he worked “most of the time” for wages, salary, commission, tips, contract or payment in kind

 In Uganda, stats are based on a labour-force survey. You are considered employed if

a) performed “some work” for pay or profit during the reference week; b) were temporarily absent from work during the reference week because of illness or leave, but were definitely going to return; and c) were engaged in production of goods for on use. “Some work” is defined as 1 hour or more during the reference week.

I checked the definitions for a few other countries and they looked rather similar – they tend to include all kinds of casual labour in their employment stats. In Uganda it was quite extreme, if you worked for 1 or 2 hours the week before the interview, you are considered “employed”. When we talk about the importance of “job creation” for development, random casual jobs for one or two hours a week is definitely not what we are talking about. The problem is that when we look at the general graph, it is difficult to make sense of the differences between some countries.

Source: Anyanwu (2013)

Source: Anyanwu (2013)

I’m not an expert of labour markets in each of these countries. But how realistic is it that Burkina Faso, Ethiopia and Central African Republic have relatively high employment rates, while Egypt, Algeria and South Africa are relatively low? Initially I thought that high employment rates could be an indicator of large informal sectors. But that seems like a partial explanation – the reality is that employment statistics are fundamentally unreliable and country-level comparisons cannot be accurate.

A recent paper by Fox and Pimhidzai looks at the problem more in-depth for Africa and particularly for Uganda. Excerpts from the abstract:

 A cursory review of employment data for low-income Sub-Saharan African countries shows both large gaps and improbable variation within countries over time and among countries, suggesting that low quality data are routinely reported by national statistics offices. Unfortunately, policies are formed and projects developed and implemented on the basis of these statistics. Therefore, errors of measurement could be having profound implications on the strategic priorities and policies of a country… [The paper] finds that estimates of employment outcomes are unreliable if the questionnaire did not use screening questions, as labor force participation will be underestimated. Likewise, surveys that use a seven-day recall period underestimate or potentially misrepresent employment outcomes, owing to seasonality and multiple jobs. […] The paper concludes that there is a knowledge gap about employment outcomes in Sub-Saharan Africa that will continue unless collection techniques improve.

So the lesson of the day is “to always be suspicious about employment stats in Africa”, especially cross-country comparisons. Thanks to  @RachelStrohm and @RowanEmslie among others for questioning the issue.