initial surplus
Before a single premium has come in or a single claim has been paid, an insurer needs a cushion of money already on hand — a buffer to absorb the bad luck of an early run of claims before premium income has had time to pile up. That starting cushion is the initial surplus. It is the height at which the surplus line begins on day one, the savings the company brings to the table before the game starts.
In the surplus process U(t) = u + c*t - S(t), the initial surplus is the constant u — the value of the balance at time zero, when no time has passed and no claims have yet occurred. A larger u lifts the whole jagged surplus line upward by the same amount, so it has farther to fall before it ever touches zero. For example, doubling u from 1 million to 2 million does not change how often or how big the claims are, but it gives every possible future path an extra million of headroom against ruin.
Initial surplus is where risk theory meets capital. The ruin probability depends heavily on u: more starting capital means a smaller chance of ever going broke. Turned around, an actuary can ask 'how much initial surplus do I need so the probability of ruin stays below, say, 0.5 percent?' — and that answer is essentially a required-capital calculation. A subtlety worth keeping straight: in this model u is real money set aside as a safety buffer, not the technical claims reserve that backs known liabilities; the surplus is what stands above those liabilities.
Two insurers write the identical book of business, but one starts with u = 0 and the other with u = 2,000,000. The claims and premiums are the same — yet the second insurer can ride out a bad early year that would have sunk the first below zero. Initial surplus buys time for premiums to catch up.
Initial surplus u is the day-one cushion; more of it pushes ruin probability down.
Initial surplus is capital above the liabilities, not the reserve that backs the liabilities themselves. Confusing the safety buffer with the technical reserve is a frequent beginner mistake.