By Romina Boarini
Building a composite index requires some steps. These are[1]: identifying the components of well-being that one wants to measure; identifying the indicators that capture best these components; carrying out imputation of missing data; normalising the indicators; deciding how to weigh and aggregate the indicators; assessing the robustness of the index through a sensitivity analysis; reflecting on the visualisation of the results.
Concerning the choice of the components, the main challenge is to come up with dimensions of well-being which are equally relevant for different countries. For doing that, Your Better Life Index relies on the OECD Measuring Progress Framework which assesses current well-being on the basis of two domains (material living conditions and quality of life) and eleven dimensions (income and wealth, jobs and earnings, housing, health status, education and skills, work and life balance, civic engagement and governance, social connections, personal security, environmental quality and life satisfaction). This framework has been chosen following theory (e.g. the Report on The Measurement of Economic Performance and Social Progress by the Stiglitz-Sen-Fitoussi Commission) and practice in many OECD Countries (e.g. Australia ABS framework to measuring progress). Its rationale is discussed in more details in the Compendium of Well-Being Indicators.
This framework and in particular the strong focus that this puts on households, inequalities, outcomes and both objective and subjective features of well-being, inspired the choice of the indicators. In addition, indicators have been identified so as to be relevant (e.g. policy amenable, easy to interpret, etc.), relying on very good data quality (e.g. most of them comes from National Statistical Offices) and comparable across the OECD countries. Finally, indicators have been discussed with National Statistical Offices of the member countries.
Despite considerable effort put in seeking the data, some indicators display missing values. We have thus estimated the missing values through standard imputation techniques. We have also tested the impact of such an imputation on the values of Your Better Life Index and concluded that imputation (which in fact concerns less than 5% of the overall data) does not affect substantially the Index.
The next step has been to normalise the indicators, i.e. expressing them in the same metrics. Normalisation is needed as indicators are originally expressed in different units (dollars, years, percentage points, etc). The normalisation technique, which is a standard one for composite indices, consists of comparing each value to the boundaries of the interval where the indicator ranges. The resulting normalised values vary between zero (the bottom performer) and one (the top performer).
Once values are normalised, they can be aggregated. In Your Better Life Index this happens in two stages. First, within each dimension, indicators are added up with equal weight. Secondly, aggregation across dimensions is done automatically and interactively when the users rate the topics. When rating the topics, users can decide to assign no rate at all (i.e. the topic is not important at all) or go up to 5 rates (i.e. the topic is very important). These rates are automatically converted into weights which vary from zero to 100%, with the constraint that the sum of weights must be 100%.
Before launching Your Better Life Index we have carried out several robustness tests and other checks, to ensure that the Index is statistically sound.[2] In particular we have tested the sensitivity of the index to the weights assigned at various levels (domains, dimensions and indicators) and to a number of other assumptions (imputation and normalisation). Overall, the results show that the Index is robust to these various assumptions.
A great deal of work has also been done on the visualisation aspects of the tool, by the external developers of the website and the OECD Communication Directorate. For instance simplicity and user-friendliness of the tool have been balanced with its level of precision. More information on this can be found at http://www.jeromecukier.net/?p=872.
No comments:
Post a Comment