Graphs can be used to great effects in publications. They can summarize trends, patterns and relationships between variables. They can illustrate and amplify the main messages of the publication, and inspire the reader to continue reading. Graphs are generally better understood and interpreted by the average reader, and therefore appeal to a wider audience. If done well, they can give readers a quick and easy understanding of the differences and similarities between women and men.
Every graph should make a point and that point can be given in the title. Nevertheless, in many publications, titles state the subject and the coverage of data in the graph. In this case, the title should start with the key word(s) of the statistics presented.
There are many types of charts. The type of chart used depends on the kind of data used in the analysis and what point the authors wish to make. Choosing the correct chart can make the difference between providing the reader with a strong message and confusing the reader.
- + Line charts
- Line charts give a clear picture of changes over time or over age cohorts that cannot easily be discovered in data tables. Time-series data that are often shown in line charts are life expectancies at birth, infant mortality, literacy rates, and labour force participation rates. In general, it is expected that advances in human development over time are reflected in declining infant mortality rates and increasing literacy rates and life expectancies;labour force participation rates are expected to respond to changes in overall market and trade conditions.
Chart 4.1, for example, shows trends in life expectancy at birth for women and men in South Africa.
It is generally recommended that charts start from zero at the y-axis of a quantitative variable, so that the differences or similarities between women and men are not distorted. However, at the same time, it is important that the comparison between women and men is facilitated. In this case all the values of life expectancy are concentrated above age 35. The focus on differences between women and men in the chart presented, allows the viewer to see how the HIV/AIDS epidemic in South Africa changed in the 1990s women’s and men’s trends in life expectancies.
Chart 4. 1 Life expectancy at birth by sex, South Africa, 1950-2010
Source of data: United Nations, 2011. World Population Prospects.
Line charts are also useful in revealing changes from one age cohort to another, in labour force participation, employment, or literacy, for example. Chart 4.2, for instance, shows age patterns in labour force participation for women and men in Chile, for two points in time. The chart illustrates three main points: (a) at all ages, labour force participation rates are lower for women than for men; (b) in the last two decades women’s participation rates increased more than men’s participation rates; and (c) in the most recent year observed, women tend to withdraw from the labour market after age 30.
Chart 4. 2 Labour force participation rate by age group, by sex, Chile, 1990 and 2008.
Source of data: ILO, LABORSTA table 1A (accessed January 2012).
- + Vertical bar charts
- Bar charts are common in presentation of gender statistics. One of the axes, usually the x-axis, is formed by a qualitative variable with distinct categories. This variable can be sex or other breakdown variables such as urban/rural areas, regions, or wealth quintiles. The other axis can represent absolute frequencies or percentages, sums or averages. Bar charts can be used to illustrate data that do not vary in magnitude too greatly.
Chart 4.3 is an example of simple vertical bar charts. It shows the percentage of women in India who have ever experienced physical violence for different categories of wealth, ordered from the poorest quintile (poorest 20 per cent of the population) to the wealthiest quintile (wealthiest 20 per cent of the population). Other examples of simple bar charts may include total fertility rate by region, antenatal care by urban/rural areas, or proportion of women married before age 18 by level of education.
Chart 4. 3 Women age 15-49 who have experienced physical violence since age 15 by wealth quintiles, India, 2005-06
Source of data: Ministry of Health and Family Welfare Government of India, 2007. National Health Family Survey 2005-06
Grouped (or clustered) bar charts present the same characteristic for two or more categories of population at the same time, thus facilitating comparisons. Often, the values of a characteristic for women and men are shown as two sets of differently colored or shaded bars side by side for each category. For example, in Chart 4.4, data on school attendance in Yemen is presented for girls and boys side by side within two categories of population,the poorest and the wealthiest quintiles. It is shown that girls have school participation rates lower than boys in both wealth groups; however, the gender gap is much more substantial in the poorest group of population
Chart 4. 4 Primary school net attendance rate for children in the poorest and wealthiest quintiles, Yemen, 2006
Source of data: Ministry of Health and Population and UNICEF, 2008. Yemen Multiple Indicator Cluster Survey 2006, Final Report
If more categories or data points need to be illustrated, the bars can become too thin and difficult to interpret. In such cases it is recommended that some dot charts are used instead of grouped bar charts. For example, by comparison to Chart 4.4, Chart 4.5 presents gender differences in school attendance for all wealth quintiles and for urban and rural areas. Chart 4.5 shows the disadvantage of girls in school participation in all groups and how this disadvantage is greater in the poorer population and in rural areas.
Chart 4. 5 Primary school net attendance rate for girls and boys by wealth quintiles and by urban/rural areas, Yemen, 2006
Source of data: Ministry of Health and Population and UNICEF, 2008. Yemen Multiple Indicator Cluster Survey 2006, Final Report
- + Stacked bar charts
- Similar to the grouped bar charts, stacked bar charts illustrate data sets consisting of two or more categories. Stacked bar charts can be used for most kinds of data but they are most effective for categories adding to 100 per cent. A common problem with stacked bar charts is that one or more segments are too short to be visible on the scale. Another problem is that more than three segments of the bar are difficult to compare from one bar to another.
Some stacked charts illustrate the percentage distribution by sex within various categories of variables, such as the share of women and men among categories of occupations. Chart 4.6 is one example of this type of stacked charts and it shows that in Viet Nam women hold only a small proportion of property titles.
Chart 4. 6 Distribution of property titles by sex of the owner and urban/rural areas, Viet Nam, 2006
Source of data: Viet Nam Ministry of Culture, Sports and Tourism and others, 2008, Results of Nation-wide Survey on the Family in Viet Nam 2006: Key Findings.
- Other stacked charts, however, can illustrate the distribution of variables within the female and male population, for example, the distribution of female and male deaths by cause of death, or the distribution of female and male employment by sector of employment. Chart 4.7, for example, shows that women’s employment in Morocco is concentrated primarily in agriculture, while men’s employment is concentrated primarily in services and secondarily in agriculture.
Chart 4. 7 Employment by aggregate sector, by sex, Morocco, 2008
Source of data: ILO – KILM. Accessed April 2012.
- + Horizontal bar charts
Bar charts can also be presented horizontally. They are often considered when many categories need to be presented, or where the categories presented have long labels. Men and women can be presented side by side for each category, as in Chart 4.8. Similar to the vertical bar charts, when the graph needs to display the sex distribution within a category and the values for women and men add up to 100 per cent, a stacked bar chart should be considered.
Chart 4. 8 Proportion of obese persons, by sex and wealth quintile, Egypt, 2008
Source of data: El-Zanaty, Fatma and Ann Way. 2009. Egypt Demographic and Health Survey 2008. Cairo, Egypt: Ministry of Health, El-Zanaty and Associates, and Macro International.
Horizontal bar charts are also ideal for showing time use data, because the left-to-right motion (in Western culture) on the x-axis generally implies the passage of time. Chart 4.9 provides such an example.
Chart 4. 9 Average time spent on care for children, sick and elderly by sex, urban/rural areas and marital status, Pakistan, 2007 (minutes per day in total population aged 10 and above)
Source of data: Government of Pakistan, Statistics Division, Federal Bureau of Statistics, 2009. Time Use Survey 2007.
Bar charts are often used to present gender statistics for different regions of a country. When there are many regions to be presented, a horizontal bar chart may be preferred. It is important that the regions considered are presented in such a way that they facilitate the comparisons between women and men within and between the regions. Ranking of the regions alphabetically is seldom a good solution. When no other dimension is the focus of analysis (such as level of economic or human development of the region, for example), it is important that the regions are presented in the graph according to the rank of values observed for women or, less frequent, for men. Ranking of the regions by gender gap may also be considered if the graph does not become too confusing.
Another way to use a horizontal bar chart is to plot against each other (extending left and right from the y-axis) two variables that are visibly correlated. Example of a pair of such variables would be the proportion of women married before age 18 and adolescent fertility rate, both disaggregated by region; or total fertility rate and women’s contraceptive use, both disaggregated by region. The two variables considered for this type of plot do not need to have the same scale.
A variation of the use of horizontal bars is the “age and sex pyramids”. Traditionally, age pyramids plot the age composition of the population of women and men as horizontal bars originating from the y-axis, using absolute number of women and men by age group. Because they use absolute numbers, they tend to emphasize the concentration of population in particular age groups. Alternatively, this type of chart can be constructed using percentages instead of absolute numbers, emphasizing the groups where women or men are overrepresented. For example, Chart 4.10 illustrates the composition of population in Swaziland by sex, age group and level of educational attainment. By comparison, Chart 4.11 illustrates, for the same country, the proportion of women and men with at least secondary education within age group.
Chart 4. 10 Distribution of population by sex, age group and educational attainment, Swaziland, 2007
Source of data: United Nations, DYB data on line. Census data sets (accessed January 2012).
Chart 4. 11 Proportion of population with at least secondary education, by sex and age group Swaziland, 2007
Source of data: United Nations, DYB data on line (accessed January 2012)
Other examples of age and sex pyramids would be foreign-born population by sex, age group and marital status; or proportion of population smoking by sex and age group.
- + Pie charts
- Pie charts are suitable for illustrating percentage distribution of qualitative variables and are an alternative to bar charts. Pie charts must always show shares that total to 100 per cent. A common error with pie charts is to show too many categories, resulting in labels that are hard to read or shares that are too narrow. When too many categories need to be compared, bar charts are more suitable.
Pie charts are best used when only one or two shares of the whole are shown for different years, different population groups or different related categories. For example, Chart 4.12 shows the percentage of women married before age 18 in urban areas as compared to rural areas. Other examples would be the share of time used by women in total time invested by women and men in various types of unpaid housework; or the share of women among managers at two points in time.
Chart 4. 12 Proportion women married before age 18 in urban and rural areas, Gambia, 2005/2006
Source of data: The Gambia MICS Survey 2005/2006 report
- + Scatter plots
- Scatter plots are often used to show the relationship between two variables. The two variables are plot against each other, in order to show patterns of their grouping. Scatter plots are also used in identifying and analyzing outliers in the data.
Scatter plots are particularly useful when many data points need to be displayed, such as in the case of a large number of regions or sub-regions of a country that cannot be easily presented in tables or bar charts. Chart 4.13, for example, shows school attendance rates for girls plot against school attendance boys for children in the states of India. The dots that are close to the diagonal represent the states where girls and boys have similar school attendance rates. This is the case for most of the states in India. However, there are a few exceptions. A number of states with generally lower school attendance have higher rates for boys than for girls. These particular cases may be highlighted on the graph.
Chart 4. 13 School attendance rates for 6-17 years old by sex and state, India, 2005-06
Source of data: India Ministry of Health and Family Welfare Government of India, 2007. National Health Family Survey 2005-06, vol.1.