7 Appendix: Extended Coverage Analysis

This appendix includes a range of options and definitions for classifying indicator coverage as discussed in the section on coverage analysis. Throughout our research we discussed several options for thinking about clusters of indicators, and the complete list, along with their formulas and definitions, are included here for reference. Several of these approaches have strong similarities to the ones that appear in the coverage analysis section, but with slightly different or more precise definitions that may be useful in future research.

7.1 Country Coverage

Country coverage is defined as follows,

\[\begin{equation} C_j = \frac{1}{y_f - y_i +1} \sum_{y = y_i}^{y_f}{n_{j}^{y}} \tag{7.1} \end{equation}\]

Basically, it is the the average of the number of countries for which there is information available in indicator \(j\) over the period \(y_i-y_f\). In this case, \(n_{j}^{y}\) is the number of countries with information available for indicator \(j\) in year \(y\). \(y_i\) and \(y_f\) refer to the initial and final years of the analysis, which at the time of this publication they are 2000 2019, respectively.

The table below is an expanded version of the table presented in section 2.1. Indicators are organized from the lowest to highest country coverage.

7.2 Improvement over time

In section 2.3 we present how some indicators have presented a remarkable improvement over time. In order to find these countries systematically and taking into account that The coverage growth in many indicators is inconsistent and fluctuating over time, we estimate the average growth in the number of countries \(\hat{G_j}\) in indicator \(j\) (this is different from \(C_j\) in equation (7.1)),

\[\begin{equation} n_{j}^{y} = \alpha + G_jy + e \tag{7.2} \end{equation}\]

Where \(n_{j}^{y}\) refers to number of countries with information available for indicator \(j\) in time \(y\). As \(n_{j}^{y}\) could be zero for any \(y\), we penalize the value of \(\hat{G_j}\) by the number of years in which they have no data at all, \(p_j\),

\[\begin{equation} p_j = \frac{Q_j}{y_f - y_i +1} \tag{7.3} \end{equation}\]

where \(Q_j\) refers to the number of years for which there is at least one country available in indicator \(j\). \(y_i\) and \(y_f\) refer to the initial and final years as in equation (7.1).

Most indicators have improved their coverage over time, but some of them show a remarkable improvement. The table below shows the average growth of number of countries in each indicator penalized by the number of years in which they have no data at all. The higher the number in column Avg. growth in coverage (penalized) the more consistent the indicator has been on increasing country coverage since 2000.

7.3 Remain stable over time but with middle coverage

Stability of country coverage over time can be seen as the standard deviation of the number of countries available for each indicator in each year. The lower the standard deviation, the more stable the country coverage of the indicator is.

The country coverage of some indicators have remained relatively stable over time. Among the 7 indicators that have not changed their country coverage over the period of analysis, the average of countries covered is 187. These indicators are Employment to population ratio, ages 15-24, total (%) (modeled ILO estimate), Vulnerable employment, total (% of total employment) (modeled ILO estimate), Labor force participation rate, total (% of total population ages 15-64) (modeled ILO estimate), Ratio of female to male labor force participation rate (%) (modeled ILO estimate), Unemployment, youth total (% of total labor force ages 15-24) (modeled ILO estimate), Unemployment, female (% of female labor force) (modeled ILO estimate), and Unemployment, total (% of total labor force) (modeled ILO estimate).

The table below shows how stable each indicator is over time and the average number of countries covered in each year.

7.4 High coverage and sudden decline

The remarkable coverage of some indicators in the previous decade suddenly declined and has remained either null or very low since then. The table below shows the year in which the country coverage of each indicator plummeted and the number countries dropped in that year.

7.5 Intermittent and Lumpy coverage

Some indicators have intermittent coverage with excellent coverage in some years and no coverage at all in others. Other indicators are very lumpy in the sense that the number of countries covered varies up and down from year to year. Following Syntetos, Boylan, and Croston (2005) and Williams (1984), we calculated the Average coverage interval and the Lumpiness index for all indicators.

The Average coverage interval indicates the average number of years between successive years with country coverage. For example, if the Avg. coverage interval is 24, it means that in average every four years this indicator covers at least one country.

\[\begin{equation} aci_j = \frac{Y}{Q_j} \tag{7.4} \end{equation}\]

Where \(Y\) is the total number of years in the period of analysis and \(Q_j\) is the number of years for which there is at least one country available in indicator \(j\), exactly the same as in equation (7.3). The higher the \(aci\), the more intervals with no information the indicator has.

The Lumpiness index shows how lumpy the indicator is. The higher it is, the more lumpy the country coverage is. It is calcualted as follows,

\[\begin{equation} L = \frac{cv(n_{j})^2}{aci} \tag{7.5} \end{equation}\]

Where \(cv(n_{j})\) is the coefficient of variation of the number of countries avialable over time in indicator \(j\) and \(aci\) is the derived in equation (7.4).

The table below tries to characterize all the indicators by their level of intermittentness and lumpiness. The No. years covered column shows the number of years with at least one country covered during the studied period. The No. intervals with no data is the number of intervals in which no country is covered. The higher this number the more intermittent the indicator is. The Avg. coverage interval indicates how often the indicator has information or how intermittent the information of this indicator is. The last columns contains the Lumpiness index.

References

Syntetos, A A, J E Boylan, and J D Croston. 2005. “On the Categorization of Demand Patterns.” Journal of the Operational Research Society 56 (5): 495–503. https://doi.org/10.1057/palgrave.jors.2601841.

Williams, T. M. 1984. “Stock Control with Sporadic and Slow-Moving Demand.” Journal of the Operational Research Society 35 (10): 939–48. https://doi.org/10.1057/jors.1984.185.