Quantifying Global Public Procurement: Our Methodology

The Open Contracting Report ‘Global Procurement Spend‘, is to our knowledge, the most comprehensive study of global public procurement yet. We are proud to be the data partner for the project, analysing more than 7 million documents to quantify the $12.9trn size of public procurement, globally.

So, you’re probably wondering how we put a number on it?

Our methodology is rigorous and we work hard to ensure our data is the best it possibly can be. Our team stay on top of the data, 365 days of the year, every year.

In order to identify with confidence with a global figure, we take our data and implement a three-step process.

1.Collating information

For the first step, we collect the procurement spend from sources such as the World Bank, IMF, OECD and government sources. We then add a range of national metrics tracked and maintained by the World Bank. These include GDP, GNI, Military Spend, Tax Revenue, Total Debt Service, Statistical capacity, HCI.

2.Establishing a relationship

We then look at the relationship between procurement spend and the other categories from the World Bank by testing it using t-test correlations. We need to compare larger variables such as GDP with smaller variables such as statistical capacity and Human Capital Index so smaller, normalised values were used. This means that all values were placed on a 0-1 scale in line with the smallest variable: giving us the statistical capacity to prevent very large values from distorting the relationships. The main 4 variables that showed strong, statistically significant correlations with procurement spend were GDP, GNI (PPP), military expenditure and revenue.

The United States and China were left out for two reasons. First, their economies are substantially larger than those of the rest of the world. Second, the scale of their economies means that their unique attributes are quite likely to distort our model. The correlations increased when the two outlier countries, the United States and China, were left out.

Owing to the limited sample size (66 countries with sourced data), only the strongest correlations (of over 0.8) were chosen to build the model. This is good practice because, with more variables, we need a larger sample size to maintain confidence in any relationship established.

3.Creating a model

From there, we start the data wrangling and create a model. We begin by banding countries on their GDP. The reason for starting here is to provide a more refined model of spend that better reflects inter-country difference and captures like for like economies. GDP had the strongest correlation with Procurement Spend.

Countries were banded by ‘large’, denoting any country with GDP higher than $1.5trn, ‘medium’, GDP’s between $1.5trn and $100bn and ‘small’, any GDP less than $100bn. These bands each created a statistically significant model.

For each band, a regression model based on the existing dataset was created and applied to countries where procurement spend is not readily available. This created a ‘best fit’ number. The number was totalled and, with the exception of China (the figures of which are published for 2018), increased by 1.905%.

Why 1.905%? Because it is the amount of growth in the world’s GDP between 2017 and 2018.

As a result, we are able to determine a total public procurement spend figure of $12.9trn per annum.

Variables

It’s important to remember that there are a number of variables that we need to consider and make adjustments to, through this process.

For example, China and India are two of the largest contributors to procurement. For India, volumes of public procurement vary because procurement happens at multiple levels, many of which are yet to use e-procurement and/or to publish contract award data.

To err on the side of caution, we chose the most conservative estimate: the lowest band of 20% GDP. For China, their public procurement spend at over $4trn takes up 30% of the model, and should be viewed with caution. This is because of the sheer scale of China’s figures, so the effects of any anomalies could be greater than a small economy. To mitigate this, we used official Chinese sources or credible sources in lieu of official estimates.

We also put in place some broad caveats. For example, for many countries, recent data was not forthcoming. We then used what we know of procurement spend patterns in OECD countries where procurement spend as a percentage of GDP remains constant and unchanged.

If you’d like to chat to us about our data capability, and how it could help you or your business, get in touch.