Are there biases in the data?
Answer
Good question! Most private investment data is contributed data and as such is likely to suffer from multiple biases:
Selection bias
Survivorship bias
Backward-looking bias
Survivorship bias in particular (only observing 'winner' or the best investments) is one of the most important issues when building an index of the risk and performance of infrastructure investments. Our research paper explains why contributed indices tend to suffer from bias.
Scientific Infra & Private Assets seeks to avoid these biases by generating its Index (or Sample) Universe using a bottom-up approach with the specific aim of building a representative sample of the investable universe: first using country or market selection rules, followed by company selection rules.
As a result, the companies are included in the universe over time without pre-judging whether they will do well or badly, survive or default, or go bankrupt.
Things to consider
The indices are constructed from the bottom up, with a representative set of 700+ infrastructure companies and 2,000+ private debt instruments spanning over 20 years.
This universe also includes firms that have faced problems and defaulted or even gone bankrupt; for example, the nine Spanish toll road companies that went into administration in 2012-2013 as well as the UK power plants which emerged from their PPA during a very unfavourable price environment and subsequently filed for bankruptcy. They have also included projects that experienced delays, lower-than-expected traffic and even developer bankruptcy.
As a result, some of the risk and low or negative returns reported in the Scientific Infra & Private Assets indices are more representative of what the average investor might expect from this asset class.