Where does run collide?

A situation in which a large number of a bank's depositors try to take out their money at the same time because they fear the bank will fail, so that the rush of withdrawals can drain its cash and make the very failure they feared come true. Diamond and Dybvig's 1983 Bank Runs, Deposit Insurance, and Liquidity modelled this as an economy where a bank turns illiquid long-term assets into money-like deposits, leaving both a calm outcome and a self-fulfilling panic possible for the same bank. The referent is a collective withdrawal of money driven by fear, measured in deposits pulled out, and it has nothing to do with a streak of like outcomes in data, a batch of goods made in one setup, or a fast passage of notes.[1]

Unit

One episode of mass withdrawal at a bank, scaled by deposits pulled out

Recovery means

A run can be halted by deposit insurance, a lender of last resort, or suspending withdrawals, each of which removes the incentive to rush

Measured by

Tracked by the volume and speed of deposit outflows and the share of depositors trying to withdraw

A maximal block of consecutive, like outcomes in an ordered series: in a string of coin tosses, for example, a stretch of heads uninterrupted by a tail is one run, and the next tail begins another. Mood's 1940 The Distribution Theory of Runs worked out how many such blocks, and of what lengths, to expect if the series were random, which is the basis of the runs test used to judge whether an ordering is random or instead shows clustering or alternation. The referent is a streak of matching results counted inside a data series, and it is unrelated to a panic of bank withdrawals, a quantity of product made in one setup, or a rapid figure of notes.[2]

Unit

One uninterrupted block of like outcomes within an ordered series

Recovery means

Not applicable

Measured by

Counted as the number and lengths of runs in a sequence, compared with the distribution expected under randomness