BDO International Business Compass

International Location Index for the Medium-Sized Companies

Compensation of Missing Data

Closing all gaps in the data set is an essential requirement in order to calculate an index. However, despite careful research and consultation of various sources, a few observations for individual indicators had to be imputed, i.e. statistical methods are used to compensate for missing data. 

In the first case, there were gaps in the unemployment rate for all countries of the data set. Therefore, the missing values for Burundi, Chad, the Comoros, the Republic of Congo, the Democratic Republic of Congo, Ivory Coast, Eritrea, Gambia, Guinea, Guinea-Bissau, Mozambique, Rwanda, the Solomon Islands and Togo had to be determined per regression. For this purpose, the employment rate of the male population was utilised as an explanatory variable. The male employment rate has a negative correlation with the unemployment rate of a particular country.[1] Only the male part of the workforce was considered in order to exclude distorting factors such as the restriction of or a ban on female employment on religious grounds. In addition, dummy variables control the different regions.

The missing values in the remaining cases of data gaps were replaced by the average value of the variable for the relevant region.[2] The values for the "population growth" variable of Taiwan and Zimbabwe were supplemented. "Investment rate" data for Iraq, Laos, Liberia and Samoa are compensated. The values for the "public debt" variable of Afghanistan, Algeria, Kosovo, Samoa, Timor-Leste (East Timor) and Vanuatu were calculated. The value for the "per capita FDI-inflows" variable of Taiwan was completed by the regional average. The "per capita household expenditure" of Afghanistan, Bhutan, Bosnia and Herzegovina, Iraq, Kosovo, Liberia, Myanmar, Nigeria, Samoa, São Tomé and Príncipe, Sri Lanka, Taiwan and the United Arab Emirates were compensated with the relevant average value. The values for "health" and "education" of Kosovo and Taiwan were supplemented in the same way. Average data had to be added for the "commercial, trade and investment freedom" data as well as the "labour freedom" data of Afghanistan, Brunei Darussalam, Iraq, Kosovo and Sudan. The data for "infrastructure" of Barbados, Belarus, Belize, Brunei Darussalam, Burundi, the Cape Verde Islands, the Central African Republic, Equatorial Guinea, Kosovo, Lesotho, Malawi, Mauritania, Morocco, Samoa, São Tomé and Príncipe, St. Lucia, Surinam, Swaziland, Taiwan, Timor-Leste, Trinidad and Tobago, Vanuatu and Zimbabwe were supplemented. The observations of the "aggregate tax rate" variable of Barbados, Libya, Malta, Myanmar, Turkmenistan and Taiwan are compensated. Only data up to and including the year 2003 were available for the "market potential" variable. The missing values up to 2010 were interpolated. Initially, the data were logarithmised, then updated and finally exponentiated. This ensures that the interpolated values are always positive. Lastly, the missing values for Kosovo, Montenegro and Timor-Leste had to be replaced by the average value of the region.[3]



1 Spearman's rank correlation coefficient = -0.4137.

2 The substitution by the measure of central tendencies of the arithmetic mean is a commonly used method of singular imputation. One has to bear in mind, however, that adding missing values with the average value of the sample or of a sub-sample leads to a distortion in the indicator with a systematic underestimation of the variance (cf. OECD (2008), 56). The variance is underestimated, because application of the central values, in this case the mean value, reduces the dispersion of the variable.

3 Appendix F illustrates three examples of "market potential" progression over time. Figure 18 shows the typical progression for Belgium, Hong Kong, Philippines, Thailand, Uganda and the United States. In these cases, the logarithmic "market potential" follows an approximately linear trend. Figure 19 illustrates the prognosis for Angola, Argentina, Romania, São Tomé and Príncipe, South Africa and Zimbabwe. In these cases, wider fluctuations are observed over time. Figure 20 shows an example of countries whose "market potential" was subject to considerable fluctuations over the available years of the time series. In the case of Iraq, the first and second Gulf War caused these fluctuations, but the Iraq War only features in the 2003 value. All the others are predicted values. Values are available only from the early 1990s onwards for Russia and Uzbekistan, two former Soviet States. In the past, Yemen has always been subject to armed conflict, causing considerable fluctuations.