A recent submission to a Senate inquiry by the Centre for Public Integrity claimed that in the decade to 2023 companies making political donations “were 2.49 times more likely to win procurement contracts than non-donors”, and that the value of contracts won by donor companies was on average 4.4 times the value of contracts won by non-donors. Centre Director Geoff Watson SC told The Guardian that “the people who are making these donations are not fools and are doing it for a reason. The data demonstrates that the donations achieve their purpose because they’re receiving more government work. It’s as simple as that.” Spoiler alert: it’s not.

When I read the Senate submission from the Centre for Public Integrity, two issues stood out for me: how did the Centre get hold of the data to support the conclusion that donors are 2.49 times more likely to win contracts? and how does the Centre imagine donor status practically affects tender outcomes?

The data issue

The submission stated that the Centre’s research drew on donations data from the Australian Electoral Commission and contract data from Austender. Together this data may support conclusions about the number of contracts awarded to donor companies as compared with non-donors. But it cannot support conclusions about the relative likelihood of winning contracts, because it does not contain data on unsuccessful attempts to win contracts.

To draw a sporting analogy, the South Melbourne/Sydney Swans has won five VFL/AFL premierships, while Port Adelaide has won one. But this doesn’t mean the Swans are five times more likely to win, because it ignores the fact that the Swans have competed for 127 seasons, and had 127 opportunities to win, while Port have only competed for 27 seasons. Adjusting for opportunities to win, the success rate for both teams is a little under four per cent.

A conclusion that donor companies are more likely to win contracts implies that their success rate is better than non-donor companies. Measuring success........

© Pearls and Irritations