As stated, in hypothesis 1 I have used the annual unemployment rates of the ILO which are included in the WDI data set collected by the World Bank. More precisely I have used modelled ILO estimates of the total unemployment rate, which are expressed as a share, or percentage, of the unemployed in the total labour force. Following ILO the unemployed are defined as “members of the economically active population who are without work but available for and seeking work, including people who have lost their jobs or who have
voluntarily left work”. These modelled ILO estimates of the unemployment rates are based on household labour force sample surveys. The figures include nationally reported and imputed data. According to the data description this data which I will use is “harmonized to account for inconsistencies resulting from data source, definition, reference period, coverage, age group, and collection methodologies”, and thus the figures should be comparable between countries and through years (World Bank 2015b).
For hypothesis 2 I have found governorate level unemployment data for one or more years between 2006 and 2010 from five of those North African Arab countries which comprise the study area of research question I. This data is drawn from national statistical authorities. I have used data for 2010 if it has been available, in other cases I have used the latest year available, however 2006 being the earliest year I have accepted. As mentioned, this governorate level unemployment data will be compared to protests and the level of organised violence between 2006 and 2011. Because the idea is partly to observe whether the
unemployment rate before the Arab Spring could be used to predict where the protests and violence took place during the Arab Spring, I have not used governorate level unemployment data of 2011 even if it has been available. Because I want to know if those governorates within a country where the unemployment rate is higher have seen more protests and
organised violence, I will not use the actual unemployment rates as such. Instead I will divide the governorates of every country into four categories, or in other words into quarters. The highest quarter will include the governorates with the highest unemployment rates within a country, the governorates with the lowest unemployment rates comprising the lowest quarter while the two quarters in the middle respectively include the governorates with intermediate unemployment rates.
Map 3. Unemployment by governorate in Morocco in 1999 and 2004.
It can seem dubious to use a timespan of six years when studying hypothesis 2, as the unemployment rates have fluctuated during that time. Morocco is the only one of these
1999 2004
Lowest quarter Highest quarter
Source of data:
High Commission for Planning of Morocco 2015
500 km
N
countries where the governorate level unemployment rates are available for every year for a relatively extensive period of time, from 1999 to 2013. I have calculated the unemployment rate quarters described above for each of these years. Map 3 presents the quarters for the years 1999 and 2004, which have the same six years difference as 2006 and 2011, the first and last years of the timespan used when studying hypothesis 2. From 1999 to 2004 the national unemployment rate in Morocco changed from 13.9% to 10.8% according to modelled ILO estimates. The national statistical authority of Morocco gives the same figures for 1999 and 2004 (World Bank 2015a). This change of 3.1 percentage points is more substantial than between any other six years from 1999 to 2013. However only in three out of 14 governorates the unemployment rate quarter has changed. In one governorate the change has been two quarters (from second lowest to highest) and in two governorates the change has been one quarter.
I have observed all possible pairs of two years with six years intervals (the first pair consists of the years 1999 and 2004, the second of the years 2000 and 2005 and so on) and all governorates of Morocco; there are 140 pairs of two figures, the first figure representing the unemployment rate quarter in a given governorate of Morocco in one year and the second figure representing the unemployment rate quarter in the same governorate six years later.
Consequently, during six years, in 68% of the governorates the quarter has remained the same, in 30% there has been a change of one quarter, in other words the quarter for the latter year is one higher or lower than for the first year. And in 2% of the governorates there has been a change of two quarters. In none of the governorates has the change been three
quarters—a change of four quarters is impossible, a jump from the first to the fourth quarter or vice versa equals a change of three quarters. I think that based on these calculations, the timespan of six years should not cause considerable problems; in general the patterns of the unemployment rates within a country remain similar during six years even if the national unemployment rate increases or decreases. Those governorates which have high
unemployment rates within the country in one year have higher unemployment rates within the country also six years later, the same applies for governorates with lower unemployment rates. Thus one additional benefit from using quarters instead of the actual unemployment rates is that yearly differences in the unemployment rates between countries will not have a great impact on quarters based on the unemployment patterns within a country. Lastly, since I combine data from different years, using the actual unemployment rates might give an
impression that the data is more accurate and comparable than it actually is.
The governorate level unemployment data I have used is in Morocco for the year 2010,
in Mauritania and Tunisia for 2008, and in Egypt and Libya for 2006 (National Office of Statistics of Mauritania 2009: 158; National Institute of Statistics of Tunisia 2010: 146;
Central Agency for Public Mobilization and Statistics of Egypt 2014; High Commission for Planning of Morocco 2015; Knoema 2015). The data of Libya is downloaded from Knoema, an online data catalogue, which collects openly accessible data. Knoema links to the original web page of the General Authority of Information of Libya, but I have not been able to access this original web page. In other cases than this I have found that data available at Knoema is consistent with the data from original data sources. In this data of Libya the unemployment rates were not originally included, but I have calculated the rates because the data included numbers of people employed in different ways (employee in a foreign company, employee in government, employer...) and a number of people seeking work. In the data provided by the Egyptian national statistical authority the unemployment rates were calculated based on the same logic and categories: the unemployment rate was the number of people seeking work divided by the sum of those employed and those seeking work.
Data about urban and rural unemployment rates are available for one or more years between 2006 and 2010 from five countries, Algeria, Egypt, Mauritania, Morocco, and Sudan (Assaad 2009: 21; National Office of Statistics of Mauritania 2009: 158; Nour 2011: 16;
National Office of Statistics of Algeria 2014; High Commission for Planning of Morocco 2016). In Egypt and Libya the figures are apparently based on census data. In other countries, finding out how exactly the national statistical authorities have gathered the governorate level unemployment data is not an easy task.
Because we in the first research question ask whether the unemployment rate is connected to the number of protests or the level of organised violence, there should be no problems with the validity of the indicator. We simply use the unemployment rate to measure the unemployment rate. When studying governorate level unemployment, the quarters of the unemployment rate are used instead of the actual unemployment rates. There are some disadvantages in doing so: quarters do not tell how wide the variation is between the lowest and the highest unemployment rates, in addition it can be that the dividing lines between quarters do not settle into natural breaks, for example it might be that the unemployment rate in the first governorate of the second quarter is very close to last governorate of the first quarter and clearly apart from other governorates of the second quarter. But still, taken
together these issues and advantages of using quarters discussed above, I think the advantages outweigh disadvantages. It can be seen that the fact that the data about the governorate level unemployment rates and urban and rural unemployment rates are from differing years affects
the validity. However, as shown with the data of Morocco, this should be a minor issue. In general, the unemployment rate or quarter is a valid indicator taking into account our research question.