Page History

 Introduction

Classifying countries by their development is difficult. One has to determine what to measure that is related to development and then measure it. There are then costs and practicality concerning: what can be measured, how useful it is, how pervasive it is, how well defined it is, how it changes over time, whether one is measuring the same thing for each country, and the cost of measuring. Various organizations such as the ITU, UNDP, CIA, World Bank etc. have come up with Indices  based on measured items such as life expectancy, GDP, literacy, phone lines, Internet penetration etc. (see below). These measures take time to gather and so are often dated and only available, if at all,  at widely separated intervals.

...

It is seen that there is a moderate to strong correlation R ² ~ 0.58. Similar correlations (R ² ~ 0.52) are seen when one compares PingER losses and jitter vs. GDP/capita.  Stronger correlations are obtained with development indices that are more technology or Internet related. For example if we correlate the PingER performance measurements (jitter, loss, throughput) with the GDP/capita and also with one of the more recent and extensive of the indices, namely the ITU's Digital Opportunity Index(DOI) then we get the following R2 R ² values. below Below are seen a table of R2 R ² for the correlations of pingER measurements with DOI and GDP/cap, followed by the actual scatter plots.

The advantage of these automated methods is that they are not subjective, they are available at regular intervals so the derivative with time is relatively easy to track. The disadvantage is that they are only measure one component of development and care has still to be taken to understand the figures and eliminate false readings. Given the strength of the correlation between the PingER performance measurements and those of GDP/capita and other development indices this is a rationale for pursuing this approach. This document outlines various of the survey based Index classification methods and then goes on to compare them with each other and the PingER measurements.

PingER Measurements 

The PingER project has been described elsewhere. Basically it uses the ubiquitous Internet ICMP echo request/response ping facility to measure or deduce metrics such as  Round Trip Times (RTT), jitter, loss, and reachability from about 40 monitoring hosts to over 600 monitored (remote) hosts in about 130 countries. As part of this project we recognized the need to extend the measurements to more countries. This meant adding some more countries. First we identified 84 countries appearing in the various indices that had no hosts monitored by PingER. We addressed categorized these as follows:

• Where no No measurement was made since its performance was not significantly diffferent from adjacent countries (this . This was particularly so true in Europe. To make easier comparisons in Europe where we added Sweden, Gibralter, Macedonia, Faroe Islands, Austria, Andorra, Bulgaria and Belgium);
• Where previously there had been little interest, in particular Greeenland;
• Mike Jensen gave us a list of hosts in 18 African countries where For many counties in Africa it was hard to find reliable hosts that did not block pings and were really located in the country as opposed to a proxy elsewhere. Mike Jensen gave us a list of hosts in 18 African countries where we previously had no PingER remote hosts. Of these Based on his list we were able to extract 13 hosts in 8 countries that appeared to be in the country and also responded to pings. These (Guinea-Bissau, Sierra leone, Seychelles, Mauritius, Liberia, Gambia, Swaziland and Djibouti) were added.

Another concern was the fact that the most comprehensive/complete set of measurements was from the SLAC site which measured over 600 hosts. Most other monitors, monitor 200 or fewer hosts. Thus there is a bias that measurements that rely on distance (e.g. RTT) will have better performance if they are close to SLAC. With the current version of PingER we can centrally maintain a list of Beacons (remote hosts that are monitored by all monitor hosts) and automatically update the monitors with the current Beacons on a daily basis.  There are competing concerns regarding increasing the hosts monitored. Each monitor/remote pair adds an extra 100bits/s to the network. Some countries with monitors, such as Palestine have limited bandwidth available. Thus though we need to increase the Beacon list, we need to do this carefully so we do not abuse monitors or remote hosts in countries with poor connectivity. Therefore, for each country with reasonable connectivity one host is selected that is reliable (based on previous PingER measurements), and represents the country. In addition for sites with limited connectivity we will restrivy the ping sizes to 100 Bytes rather than both 100 & 1000 Bytes, i.e. a reduction in traffic by a factor of 10. The idea is to come up with a list of about 120 Beacons covering most countries, thus roughly doubling our current list.  For more on this see PingER Beacons Expansion.

Gross Domestic Product per capita

• We also obtained hosts and contacts from

We also obtained a list of 193 Ookla Speedtest servers in 68 countries together with their and locations (latitude/longitudes). From these we added 135 remote hosts in 66 countries  (excluding hosts that did not ping and hosts in Canada and the United States, where we had adequate coverage for our purposes) to the PingER database. Finally we obtained about 10 hosts in Africa from contacts following the presentation at the  "2nd IHY-Africa Workshop" 11-16 November 2007, Addis Ababa, Ethiopia, and the presentation at the "Internet & Grids in Africa: An Asset for African Scientists for the Benefit of African Society"

Another concern was the fact that the most comprehensive/complete set of measurements was from the SLAC site which measured over 600 hosts. Most other monitors, monitor 200 or fewer hosts. We found biased results for the measurements which rely on distance (e.g. RTT) will have better performance if they are close to SLAC. With the current version of PingER we can centrally maintain a list of Beacons (remote hosts that are monitored by all monitor hosts) and automatically update the monitors with the current Beacons on a daily basis.  There are competing concerns regarding increasing the hosts monitored. Each monitor/remote pair adds an extra 100bits/s to the network. Some countries with monitors, such as Palestine have limited bandwidth available. Thus though we need to increase the Beacon list, we need to do this carefully so we do not abuse monitors or remote hosts in countries with poor connectivity. Therefore, for each country with reasonable connectivity one host is selected that is reliable (based on previous PingER measurements), and represents the country. In addition for sites with limited connectivity we will restrict the ping sizes to 100 Bytes rather than both 100 & 1000 Bytes, i.e. a reduction in traffic by a factor of 10. The idea is to come up with a list of about 120 Beacons covering most countries, thus roughly doubling our current list.  For more on this see PingER Beacons Expansion

 Indices Used By The Report

Gross Domestic Product per capita

The gross domestic product or GDP is a way of measuring the size of a region's economy. It is usually normalized by dividing by capita. It is often compared with the purchasing power parity (PPP) of the currency relative to the US$.  The terminology for GDP is changing to Gross National Income GNI, see the World Bank's Data and Statistics web page. The gross domestic product or GDP is a way of measuring the size of a region's economy. It is usually normalized by dividing by capita. It is often compared with the purchasing power parity (PPP) of the currency relative to the US$.  The terminology for GDP is changing to Gross National Income GNI, see the World Bank's Data and Statistics web page. There are measures from the World Bank and the Central Intelligence Agency among others, we are using https://www.cia.gov/library/publications/the-world-factbook/rankorder/2004rank.html

Human Development Index (HDI)

The UNDP  Human Development Index  report of 2006 was compiled on data from 2004 and covered 175 UN member countries (out of 192). It is a comparative measure of life expectancy, literacy, education, and standards of living for countries worldwide. More specifically:

...

It is a standard means of measuring well-being, especially child welfare. It is used to distinguish whether the country is a developed, a developing, or an under-developed country, and also to measure the impact of economic policies on quality of life. The index was developed in 1990 by Pakistani economist Mahbub ul Haq.

Digital Access Index (DAI) 

The Digital Access Index (DAI) from the ITU has data from  1995   to 2003. It combines eight variables, covering five areas, to provide an overall country score. The areas are availability of infrastructure, afordability of access, educational level, quality of ICT services, and Internet usage. The results of the Index point to potential stumbling blocks in ICT adoption and can help countries identify their relative strengths and weaknesses.

Network Readiness Index (NRI)

The Network Readiness Index (NRI) was used in the World Economic Forum's Global Information Technology Report 2007- 2007. It covers about 120 countries. It rests on three main subindexessub-indexes:

• the The presence of an ICT-conducive environment in a given country by assessing a number of features of the broad business environment, some regulatory aspects, and the soft and hard infrastructure for ICT;
• the The level of ICT readiness and propensity of the three main national stakeholders---individuals, the business sector, and the government; and
• the The actual use of ICT by the above three stakeholders.

Technology Achievement Index (TAI)

The United Nations Development Programme (UNDP) introduced the Technology Achievement Index (TAI) in 2001 to reflects a country's capacity to participate in the technological innovations of the network age. It contains data from 1995-1000 andf 2000 and covers 72 countries. The TAI aims to capture how well a country is creating and /diffusing technology and building a human skill base. It includes the following dimensions: Creation of technology (e.g. patents, royalty receipts); diffusion of recent innovations (Internet hosts/capita, high & medium tech exports as share of all exports); Diffusion of old innovations (log phones/capita, log of electric consumption/capita); Human skills (mean years of schooling, gross enrollment in tertiary level in science, math & engineering).

Digital Opportunity Index (DOI) and Opportunity Index (OI)

In 2006 the ITU submitted the Digital Opportunity Indexreport for 180 economies worldwide. It is related to the ITU's earlier DAI. It is based on 2004/2005 data. It uses 11 indicators each normalized for population or homes. the This include coverage by mobile telephony, Internet tariffs, # computers, # fixed line phones, # mobile subscribers & # internet users. 

A related Index is the ICT Opportunity Index (OI). This is based on many of the indicators used by the DOI but adds TVs, literacy, estudent e-student enrollment & International bandwidth. It has indices for 1996-2003 and covers 139 economies. This and the DOI are the only two indices specifically endorsed by the WSIS for use in the approved evaluation methodology.

Due to the DOI being one with the most recent results, having a large coverage and being blessed by the WSIS, we tend to prefer this one at the moment.

Corruption Perception Index

Since 1995, Transparency International has published an annual Corruption Perceptions Index (CPI)1 ordering the countries of the world according to "the degree to which corruption is perceived to exist among public officials and politicians".2 The organization defines corruption as "the abuse of entrusted power for private gain".3

The 2003 poll covered 133 countries; the 2007 survey, 180. A higher score means less (perceived) corruption. The results show seven out of every ten countries (and nine out of every ten developing countries) with an index of less than 5 points out of 10countries) with an index of less than 5 points out of 10.

Happy Planet Index

The Happy Planet Index reveals the ecological efficiency with which human well-being is delivered

The index combines environmental impact with human well-being to measure the environmental efficiency with which, country by country, people live long and happy lives. Learn about the ideas behind the HPI, how it is calculated, why we need it and what it can teach us. Below is an Excel correlation plot of the HPI vs PingER's normalized derived throughput. It is seen that there is little correlation.

Image Added

Summary

The following table summarize the indices based on their source, coverage and currentness.

Of these indices we chose to focus on the HDI since it measures human development and the DOI since it is still being developed by the ITU, it represents the technical/economic development, it has recent data and has an extensive coverage.

Correlations Between Indices 

 These are shown below:

Indices trends

One index that has data going back a few years is the ICT OI.   The OI trends for the world regions is seen below.

...

A linear fit cannot be right else at some time in the past the OI must have been negative.

Maps

Some maps of the index values are seen below:

The maps for ICT OI indicate that from 2001 to 2005: Latin America, N. Africa, India, Australia, New Zealand, South Europe, Russia, China are improving. Little improvement is seen in the Sub-Saharan Africa region.

PingER metrics

These are described in the Tutorial on Internet Monitoring and PingER at SLAC.

Normalized Derived TCP Throughput 

The normalization is to reduce the impacts of the derived throughput being proportional to 1/RTT. Thus sites close to the measurement/monitor host will have better derived throughputs.  Thus we calculate:

...

Comparisons Our focus is more on DOI because its data is more current, it is in active development by the ITU, and it covers most of the countries. Comparison of PingER parameters with the DOI are shown below:

Comparisons with GDP/capita are shown below.
 

Throughput vs International Bandwidth 

Comparisons with International bandwidth for 2005 are shown below. The differences in absolute values is to be expected since many end sites are sharing the international bandwidth. Thus the international bandwidth typically needs to be larger than the last mile bandwidth and the latter is often what dictates the overall bandwidth of a connection. Also the last mile is often congested so a typical session  will only get a fraction of it. A further  factor is that the PingER throughput is derived from the Mathis formula (TCP throughput = 8*1460/(RTT*sqrt(loss)) which assumes loss is driven by the TCP congestion algorithm whereas we are using ping to sample it.

We need to look at the outliers to see why they do not correlate well.  Possibilities that come to mind are:

...

The next largest outliers are Estonia, Australia and New Zealand. The latter 2 have been improving their connectivity recently. Since the International Bandwdith measurements are from 2005,  it may bIf be if one compares the normalized throughput for 2006 vs the international bandwidth the residuals for Australia and New Zealand are < 0.5. 

...