3.2 Forecasting Dividends
Key points
Historical time series of dividends are too erratic for infrastructure companies to allow a direct dividend forecast.
A more robust strategy is to use a model of the firm's free-cash flow to equity, which is always observable and to derive future dividends by combining this FCFE forecast with a FCF Retention Rate forecast.
A combination of systematic and idiosyncratic models allows using the entire history of dividend payout behaviour available for 600+ tracked firms with the unique cash flow trend of each company.
Forecasting dividend payouts is challenging because for many firms equity and quasi-equity payouts are not statistically tractable directly. As is well documented in the corporate-finance literature, private firms tend to have a more erratic dividend-payout behaviour than listed firms (dividends are less “sticky”), and their equity payouts can vary considerably in size and frequency. In fact, the best model of broad equity payouts across the infrastructure universe is a random march.
About realised infrastructure dividends
While some infrastructure companies pay dividends regularly, others pay dividends in irregular and more unpredictable patterns. A non-negligible subset of companies in the Scientific Infra & Private Assets database (about 10%) has never paid out a single dividend, some in more than ten years of operation. Still, these “zero payout” firms can be assumed to have a positive present value (otherwise investors would not hold them). Hence, they should not be excluded from a broad market infrastructure equity index, since they represent a certain pattern of equity payout found in the market.
Quantities of interest
Our approach to forecasting cash flows in unlisted infrastructure projects aims to minimise the multiplication of estimation errors by using the smallest number of variables possible. We focus on modelling the free cash flow to equity of infrastructure companies as a stochastic process described as a two-dimensional state vector (mean and variance). This is parsimonious.
To forecast the future stream of dividends, we follow a multi-step approach that involves revenue forecasting, Cash Flow Available for Debt Servicing (CFADS) estimation, Debt Service Covering Ratio (DSCR) forecasting, retention rate calibration and free cash flow modelling.