Actually using SWIID dataset is very tricky. The series are not always consistent and definitively not comparable across countries. You should check out the study Inequality in a Lower Growth Latin America, where we have a discussion on this and other datasets and the work we have done to build a consistent set of ginis. I should say that we were using within country variations and thus we just needed the series to be consistent within countries (this is also a challenge).

Dear Onur, firstly you should define kind of gini you wanted. Gini net mean a measurment income distribution after tax of personal income; meanwhile gini market is refering a measurment of inequality before income tax, such as original Pareto Theory of personal income distribution. I hope helpe you

hundred columns of an index may mean you have some different methodologies of measuring your Gini index. As you are not concerned with how the gini is established, it looks like questions that your database have to answer. Look the dictionary of your database.

As Guillermo answered here, anyway this database isn't well constructed, and country comparison is very complicated, once different methodologies are used in each country, which means in biases on estimations.

As far as I know, the Standardized World Income Inequality Database (SWIID) tries to enlarge the comparability of Gini coefficients of gross and net incomes, using a combination of WIDER and LIS, this latter being used as a benchmark in the calculation of Gini coefficients. SWIID gives Gini coefficients for a number of countries at (irregular) intervals. In all cases, the datasets are not comprehensive collections of Gini coefficients for a long-run analysis of inequality, as availability of data is limited. Thus, it is very important that before using the dataset, you define the aim of your analysis (e.g. inequalities in market incomes or disposable incomes?) and its time span (short-run/long-run?). Another issue is to look at the distribution underlying the calculation of the Gini coefficient (consumption - whether equivalised or not - income - and in this case which kind of income is included in Gini).

I think you can get useful information on methodology from http://dipeco.uniroma3.it/public/WP%20163%20Liberati%202012.pdf or from Liberati P.. The World Distribution of Income 1970-2009, Review of Income and Wealth Volume 61,  Issue 2, pages 248–273, June 2015.

Hello Onur,

If I may add, I have checked the SWIID data, apparently the data reflects the Gini coefficient (be it at net, or at market i.e. gross) at each percentile on the population. Which means the gini coefficient is given at lowest 1 percentile until top 100th percentile. If you are looking at one specific gini value for a country for a year, you need to take the average of all the hundred gini coefficients to get one gini value (sum them up and divide by hundred). This means you need a single gini measure for a country for a specific year because you want to compare cross-country gini coefficient. If you are looking at within country gini, you may use perhaps top 20% against bottom 20% of the gini coefficients of a specific country, for example, which again depends on your objective. I crosschecked the averaged data against WIID (which gives a single gini value for each country each year) and they indeed look similar.

Here are the excerpt I cite from Bergh and Nilsson "Do liberalisation and globalisation increase income inequality" (2010) which I believe they explained really well why the SWIID data is better than WIID and other gini data in term of comparability and availability: 

...Among the most commonly used measures of inequality are the Gini coefficients. For completely egalitarian income distributions in which the whole population has the same income, the Gini coefficient takes a value of 0. A value of 1 indicates that all incomes are concentrated in one person. Gini coefficients can be calculated in several ways: for gross income (before taxes and transfers), net income (after taxes and transfers), and consumption expenditure. Furthermore, the unit of analysis can be individuals or households. The lack of comparable Gini coefficients both between countries and over time has long been a major obstacle in inequality research. Many consider the Luxembourg Income Study (LIS) to be the best option, as it is based on reliable microdata from national household income surveys. Unfortunately, LIS data are available for only thirty countries, almost exclusively rich ones, and contain few observations from before 1990. As a second best solution, many scholars resort to the World Income Inequality Database (WIID), created by the World Institute for Development Economics Research of the United Nations University (UNU-WIDER). This is an updated and expanded version of the Deininger and Squire (1996) dataset, used by, for example, Berggren (1999). The WIID contains a large set of inequality statistics from several sources, totaling over 5000 observations from 160 countries. However, as Deininger and Squire have themselves pointed out, the observations are rarely comparable across countries or over time within a single country.

Two recent papers have attempted to handle the problem of few and non-comparable Gini measures: the Standardized Income Distribution Database (SIDD) created by Babones and Alvarez-Rivadulla (2007), and the Standardized World Income Inequality Database (SWIID) created by Solt (2008). Both the SWIID and the SIDD aim to improve data availability and comparability for cross- national research by exploiting the fact that different types of Gini coefficients display systematic relationships. The Gini coefficient of gross income is typically larger than the coefficient of net income, which in turn is larger than the Gini coefficient of expenditure. Similarly, Gini coefficients for households are typically lower than coefficients calculated on an individual basis.10 For example, Deininger and Squire (1996) recommend adding three points to net-income-based inequality observations to make them comparable with the gross-income-based observations....

Hope this helps. Cheers.