Correlations and Anomalies in World Bank Indicator Data

Transcription

1 Correlations and Anomalies in World Bank Indicator Data Hitesh Maidasani Varun Manjunatha October 11, Introduction The World Bank releases a compilation of global indicators which comprise various statistics in diverse categories such as: agriculture, health, economy, gender, labor, science and technology, etc. Under each of these categories, there is a total of 331 indicators which are considered important. Our objectives in this work are two-fold : First, we observed a few of these indicators to look for correlations between the data. Does a positive trend in one indicator lead to a positive trend in another indicator? Second, we wish to observe anomalies in time series data. All of the data available to us has been accumulated over a number of years. By using appropriate visualization software, is it possible to see major historical events with just numbers, and without any knowledge of history itself. 2 Dataset The dataset we used was the world bank indicator data [1]. Specifically, the indicators we used were: life expectancy (male), life expectancy (female), percentage of HIV positive (ages 15-49), improved sanitation facilities, primary school completion rate, net official development assistant (ODA) and aid received, per capita gross national income, population growth and inflation. 3 Spotfire Headlines 3.1 Women Live Longer than Men, Except for Countries with a HIV Epidemic By plotting life expectancy of men and women on a simple line graph, we found that in almost every country, the life expectancy of women is greater than the life expectancy of men. The precise scientific reasons for this trend are still disputed today. It is very interesting to note the exceptions to this rule. In 1

2 Figure 1: Male and female life expectancy and % of HIV positive for the year the early 90s (Figure 1), the only countries where men lived longer than women were Afghanistan, Bangladesh, Maldives and Nepal. However, in the late 90s and 2000 onwards, this has changed to Botswana, Lesotho, Qatar, Swaziland and Zimbabwe (sampled in 2003, Figure 2). While Qatar can be discounted due to its large immigrant population, the other nations in the list are those with a HIV epidemic (indicated by the yellow spikes). This was observed by overlaying the difference between male and female life expectancies with percentage of HIV positive population. For example, in Botswana in 2003, a whopping 26.6% of the population was HIV positive. observing the time series data, we have also come to the conclusion that HIV seems to affect the life expectancy of women more than men. 3.2 Better Sanitation Facilities Correlate to Greater Primary School Completion Rates It has been hypothesized that by providing clean water and adequate sanitation facilities, the condition of slums and other poverty stricken localities can be improved [2]. This follows logically, as good sanitation would provide better living conditions, healthier parents and children, and subsequently, more of them would attend school. To visualize this, we plotted a scattergram of primary school completion rate over availability of improved sanitation facilities (Figure 3). We found a strong positive correlation between the two - those countries which provided its citizens with better sanitation also had more children finishing primary school. Our visualization does not indicate a causation, but just a trend. Outliers would then indicate countries which provided good sanitation but where children did not attend school, (and vice versa). In the figure, coun- 2

3 Figure 2: Male and female life expectancy and % of HIV +ve for the year tries should aim for a place in the top right corner. However, Chad, Niger and Burkina Faso are examples of countries where neither facilties were adequately provided. Curiously, Equatorial Guinea and Maldives had similar sanitation facilities, but very different primary school completion rates. 3.3 Most of the countries which receive financial aid are poverty stricken, but Israel is a notable exception A treemap [3] over time of global financial aid reveals several interesting facts (Figure 4). In the 1960s, India was at the top of nations receiving financial aid. Ever since the presidency of Anwar Sadat in the 1970s, Egypt has received a disproportionately large amount of financial aid (almost entirely from the USA). In more recent years, especially due to war efforts, Iraq and Afghanistan have received large financial doles from the US government. We then decided to explore whether a country actually deserved financial aid or not. A good measure of the wealth of citizens of a country, is the Gross National Income Per Capita (GNI). By observing a plot of GNI vs Financial Aid, we can came to the conclusion that over the years, Israel has received a large amount of financial aid, even though it is a wealthy country (Figure 5). Also, we found that countries like Somalia, Myanmar, Burundi, Sierra Leone and Nepal, to name a few, are impoverished countries that desperately require financial assistant, but do not receive it. 3

4 Figure 3: Scattergram of Primary School Completion Rate vs Improved Sanitation Facilities in Notice the position of Maldives and Equatorial Guinea on either side of the best fit line. Figure 4: A treemap of aid received by countries in 1985 in USD. 4

5 Figure 5: A scattergram of Gross National Income per Capita vs Aid received in USD highlighting the position of Israel. 4 Tableau Insights 4.1 Population Growth as a Time Series The presence of time series data allows us to observe indicators over time. For example, population growth of a nation is an important indicator, and negative population growths correspond to drastic events happening in that region. In our population growth map, black dots indicate positive growth, while red dots indicate population decline (Figure 6). By fast-forwarding the data from 1960 onwards, we observe a decline during a coup in Equatorial Guinea in 1974, a decline in Cambodia during the Khmer Rouge reign and in East Timor during the Indonesian invasion in the 1970s. We also note the effect of the Soviet war in Afghanistan in the 1980s, the Croatian War of Independence, the Bosnian War, the Rwanda Genocide and the Kosovo War in the 1990s. Thus, we can efficiently observe historical events just using data and visualization tools. 4.2 Inflation Rate as a Time Series Similar to the previous visualization of population growth, we also plot Inflation rates as a time series (Figure 7). We notice that inflation was historically high during 1989 in Nicaragua as a result of the Contra War, and Hurrican Hugo. Around the same time, several South American countries (Bolivia, Peru and Brazil) also suffered due to high inflation but were able to rectify it with smart economic policies. The aftermath of the Rwanda Genocide (see previous section), led to civil war and hyperinflation in Zaire (Democratic Republic of Congo) in the 1990s. Countries with a Soviet Past like Georgia, Belarus and 5

6 Figure 6: A snapshot in 1993 of a time series world map of population growth. Black dots indicate strongly positive rates, while red indicate strong decline in population levels. Notice the red dots indicating Rwanda and Bosnia. 6

7 Figure 7: A snapshot of Inflation rates in Notice the effects of the Rwanda Genocide in the Democratic Republic of Congo, and high levels of inflation in past Soviet countries. Kazakhstan also suffered from inflation. One country which is synonymous with hyperinflation, but is missing from our list is Zimbabwe, most likely because of inaccurate data. 5 Pros and Cons of the Visualization Tools 5.1 Spotfire Pros: 1. Can easily and quickly create a variety of visualizations from raw data. 2. Can change from one visualization to another by pressing buttons on the menu. 3. Allows for adding multiple data sets from various sources by matching relations Cons: 1. Need to do offline data formatting in order to work with timeseries data 2. Difficult to use multiple data sources where date ranges do not match or where there is missing data entries. 7

8 3. Difficult to select labels for scatter plot markers and to modify legend labels. 5.2 Tableau Pros: 1. Very good for mapping geographical data - it can instantaneously recognize country names (and US county names). 2. Very easy to plot time series data, however, we need a Microsoft Excel plugin for this. 3. Can be easily embedded to a web-page, with interactions Cons: 1. Operations like zooming and panning over the map were clunky. 2. Difficult to convert to other visualizations, many of which are ghosted out. 3. Presents the sum of measures by default, which is totally unnecessary. 4. The color range cannot be made logarithmic. In general, data with a large dynamic range is difficult to visualize in Tableau. References [1] The World Bank. [2] Slum Networking of Indore City. Ind.pdf [3] Tree-maps: A space-filling approach to the visualization of hierarchical information structures. B Johnson, B Shneiderman, Visualization, Visualization 91, Proceedings., IEEE Conference on,