Mostly programs will either have a positive return on investment in the conventional sense (e.g. hire more tax inspectors, make more tax income),

or have an arguable, positive overall long-term return on investment hopefully, (e.g. infrastructure spending, sending smart kids to college)

or will enjoy widespread public support and provide non-monetary benefits, (e.g. public parks, libraries, maternity leave)

or will be widely accepted as a core function of government (e.g. police and military), or a morally good and virtuous act (e.g. caring for the disabled, foreign aid)

or will be the pet project of someone powerful, semi-useful, and inexpensive enough no-one cares to fight them over it (e.g. opera subsidy)

or will convince people they're one of the above regardless of the truth of the matter

or will be at risk of cancellation.

Government programs like research funding aren't measured by cash flow but rather by long term impact on GDP (or perhaps other metrics related to quality of life). For basic research in the hard sciences we can draw a direct line from grant funding in various fields to commercial products years later. When it comes to allocating a limited pool of government grant funding across fields, this shouldn't be the only factor but it has to be a factor. If nothing else this helps to ensure continued public support for government science funding because taxpayers see that their money isn't being frittered away.