This means that the money available under the ESIF has been spent by the member state. Ireland has spent 81% of its allocation for the 2014 to 2020 multi-annual financial framework, MFF, period, so as I say it is more or less top of the class.
I will move on from the key figures to the 2021 annual report and the main findings we have. As was the case for many years, we gave a clean opinion on the reliability of the accounts. We also gave a clean opinion on the revenue but we would like to flag that when we talk about revenue we talk about the revenue which is booked. We are not covering things like the customs or VAT gap here; that is not included. We do not know what it is because the gap is there but it is potential revenue which has not been collected by the EU. For the first year we gave an opinion on the expenditure of the Resilience and Recovery Fund. For the third year in a row we are giving an adverse opinion on the expenditure for the EU budget. Again we see a slight increase to 3% from 2.7% in the previous year. The adverse opinion is a result of this error being across a substantial part of our audit population. The main finding we still show is that there are weaknesses in the control system that is in place, both at Commission and member state levels. These weaknesses result in errors not being detected. Our presentation shows our audit population for 2021. The two big spending areas, as ever, are natural resources and environment, which represent almost 40% of our audit population, or around €56.6 billion. Cohesion, resilience and values come in second at 33.5%, or almost €48 billion. If you take those together you are looking at almost 75% of the EU budget. The next biggest heading is Single Market, innovation and digital, which include research and development, and they were around 10% of the budget, or €14.3 billion. There is not much difference in the trends. Our audit population in 2020 was just above that level of €142 billion at almost €148 billion, so there was no great fluctuation between the two years in terms of the audit population.
We have an error rate for the budget as a whole of 3%, up from 2.7%.
Components of this come from the different policy areas. As can be seen, Single Market innovation and digital has increased to 4.4% but the trend generally is approximately 4% in this particular area. It is mainly related to research and development expenditure and errors in that domain. There was one outlier in 2018 when a new programme was starting. It was a very particular circumstance but, generally, if we look back over the years, it is approximately 4%.
Regarding cohesion funding, we see a slight increase from the previous year, up to 3.6%. This funding is always a problematic area for us because it is mainly cost reimbursement-based expenditure, which leads to this type of error level.
In natural resources and environment, we have said that we are around the materiality threshold which is 2%. This is mainly due to the fact that the vast bulk of this expenditure is direct payments and is entitlement-based. That means that the level of error generally tends to be lower. Similar types of error contribute to this error rate, including ineligible costs, ineligible projects, procurement and state aid. Lack of supporting documents has increased, mainly due to the pandemic. It was more difficult to obtain the relevant documentation. We hope to see that element decrease again. The main component is ineligible costs, which contribute approximately 40% to our error rates.
In the main policy areas, economic, territorial and social cohesion, the audit population, which I referred to earlier, is €48 billion. The error rate is 3.6% so it is affected by material error. We tested 243 transactions for this policy area and Ireland was part of the sample. This year, a number of projects in Ireland were sampled. As can be seen, they equate mainly with the table we showed on the previous page, where it is ineligible projects, infringement of internal market rules and, again, lack of supporting documentation.
Our conclusion in terms of cohesion funding, is that we can not rely on the Commission's figures. It has its own figures, based on the work of the audit authorities in the member states and its own control work. We have audited both of them and we find that we cannot fully rely on them. For instance we were trying a different approach to cohesion funding to reduce the administrative burden on final beneficiaries so we had agreed that we would audit the same transactions that the other authorities had audited. We are still finding error rates of 3.5% to 3.6% which is a bit disappointing, given that they have been through at least part of the control system.
For Ireland under this chapter, €188 million is a lot of money but, relatively speaking, it is quite a low element of the EU funding that we receive. There was a problem and it is highlighted in the annual report. It was not a quantified error but it was an error nonetheless. We saw that there were cases where Ireland was using one method to claim the money from the Commission and a different methodology to pay the beneficiary. Obviously, they should match. This related to the use of simplified cost options. These are not a silver bullet. We encourage them because it should result in fewer errors but we still find some. It definitely has the potential to reduce administrative workload and errors, which is what we all want. This is implemented either by applying flat rates or having standard scales of the units of cost.
The next chapter is the most important one, in terms of funding, at least for Ireland, which is natural resources. Our audit population was €56.6 billion. We are saying that we are around the materiality threshold of 2% at 1.8%. We decided not to say that it is definitively below because we have sources of information that suggest this is a real minimum and it is possibly higher. That is why we say it is close to materiality rather than definitively saying it is below.
As I said earlier, direct payments as a whole are generally free from material error. They are mainly entitlement-based and, therefore, they are less prone to error. Ireland receives the ninth highest amount among member states under this budget heading. We had one error that involved the over-declaration of an area, which led to a quantifiable error. The main sources of error for this particular policy area are again, ineligible beneficiaries, inaccurate information on areas, non-compliance with procurement or grant award rules and typically administrative errors.
Ireland does not normally feature very much as a named member state. However, this year we have a number of cases. I referred to the simplified cost options in chapter 6. Under revenue, Ireland was a sampled member state. As I said earlier, the revenue is based on what is booked and we say it is free from material error. Ireland was part of the system for examining the traditional own resources, TOR, accounting and management in three member states. Ireland, in addition to Italy and Sweden, was selected on the basis of the amount of customs duties collected and the risk assessment. No significant problems were identified but shortcomings included the way Ireland manages the recovery of outstanding customs debts and delays on the part of making recovered traditional own resources available to the EU budget.
Another important policy heading and one which is becoming more relevant and pertinent is migration and border management, security and defence. The budget is €3.2 billion. Ireland was sampled in this policy area and we had a quantifiable error, which is mentioned in box 7.1. It related to the provision of emergency accommodation through a procurement procedure for asylum seekers in hotels. We audited this procedure and we concluded that there were so many issues with it that we considered that it was 100% error. The Commission is now going to visit to determine how much, if any, should be recovered.
