Moody’s did a study in October 2013 showing the pairwise asset correlations they use for rating structured finance transactions backed by infrastructure debt. The correlations are based on observed changes in credit quality for infrastructure projects.(1) Below are findings that pertain to Sequoia Economic Infrastructure Income Fund (www.seqifund.com) and other infrastructure debt fund strategies. Infrastructure asset correlations are significantly reduced further when infrastructure portfolios are invested in different regions, countries, sectors and sub-sectors.
SEQI asset correlations 1% to 8%. Economic demand-based infrastructure assets have correlations that range from 5-15% (see Moody’s extract below: Intra-sector asset correlations, Large Infrastructure, page 3, columns 2-4). By investing in different regions, countries and subsectors, economic infrastructure correlations drop to 5-8%. By further adding non-economic infrastructure assets, as SEQI can do for up to 20%, asset correlations for this segment of the portfolio drop to 1-5% (see Inter-sector correlations, page 3, column 2). Example assets in SEQI that fit in Moody’s non-economic infrastructure categories are regulated utilities, availability-based toll roads and power generation (renewables).
Single-region fund asset correlations 3% to 8%+. Some infrastructure funds focus on one region and 1-2 main sectors within that region.(2) An example would be an EU renewables fund. Power Generation Renewable infrastructure projects have correlations that range from 3-8% (see Moody’s extract below: PPP/PFI and Renewables, page 2, column 3). Many of these funds take early stage construction risk. Although Moody’s doesn’t penalize renewables for construction risk as they do other sectors such as PPP/PFI, they have said that more complicated renewables projects such as off-shore wind would be subject to higher asset correlation.(3)
Single-country fund asset correlations 7% to 20-35%. Some infrastructure funds invest all or nearly all (e.g., 90-100%) of their assets in one country and one sector and in some cases just one sub-sector. An example would be a renewables fund in the UK. As mentioned above, Power Generation Renewable infrastructure projects in the same region have asset correlations that range from 3-8%. By staying in one country, renewable correlations increase to 7-20%. By staying in one country and one subsector within renewables, e.g., solar only, correlations jump to 20%. Other single country funds invest in renewables and PPP/PFI. Here asset correlations are 7-20%. If construction risk is added to this investment style for the PPP/PFI assets, correlation jumps up to as high as 35% for this portion of the portfolio.
Diversification and alpha. Another important point is that infrastructure debt portfolios benefit significantly more from diversification than other asset classes such as traditional credit. This is due to the very low correlation between infrastructure assets, especially when the assets are well-diversified and fully operational. Moody’s estimates asset correlation in leveraged loans and corporate bonds to be higher than infrastructure and more concentrated in the 12-15% range. Our research has shown that 70-80% of the variability in infrastructure portfolio returns can be diversified away by adding 17-18 assets in different regions, countries, sectors and subsectors. A leveraged loan portfolio would require an estimated 70+ assets to achieve the same level of diversity, which means the portfolio would end up looking much more like the market and give less opportunity for positive alpha.
Conclusion. Two of the keys to SEQI’s low asset correlations vs. other infrastructure debt funds are (1) investing in different regions, countries, sectors and subsectors and (2) taking on very little construction risk. The construction industry is highly correlated with the business cycle. By introducing significant amounts of construction risk, some funds introduce large amounts of non-infrastructure, business cycle risk.
A note on the numbers: when banks or rating agencies talk about correlations they are usually referring to unconditional correlations, which are correlations when the markets are more or less stable. Some correlation studies done post the financial crisis are beginning to talk about conditional correlations, which are present at market extremes. The numbers in the Moody’s research piece should be taken as a good guide during most market conditions.
(1) “Moody’s Approach to Rating CDOs Backed by Project Finance and Infra Assets,” Oct 2013. See a partial list of these correlations below.
(2) Main sectors in the study are referred to as Industrial Sectors and are shown in blue headings in the far left-hand column, whereas sub-sectors within these main sectors are shown in the far right-hand column.
(3) Moody’s adds 15 percentage points as a penalty for construction risk in most sectors, e.g., PPP/PFI. SEQI takes very little construction risk so the 1-8% estimate above does not have this adjustment.