Voice and exit are both optimally effective when exercised by only a portion of the clientele. If expression of dissatisfaction becomes too widespread, it hampers the ability of a firm or agency to respond. (This is not as obviously true to me as the debilitating effect of a too extensive use of exit.)
For any organization, one of these forms of response to deteriorating quality will dominate, and this keeps the other response from ever reaching a debilitating level. For firms in competitive markets, exit is dominant, so a firm would already have been destroyed by exit before voice could create a burden.
Voice can be understood at first approximation as a residual of exit. Voice gets exercised inversely to the proportion of customers who leave: it depends on the inelasticity of demand with respect to quality.
In undeveloped economies with few substitute choices of goods, there is more use of voice. In deeper markets with more choices, exit tends to prevail.
Where exit is the dominant response, any use of voice will be beneficial. The optimal combined response to a decline in quality is an initially elastic decline in demand -- rapid exit -- followed by by an inelastic decline that gives scope to the use of voice.
Voice can also be conceived of as an alternative to exit. In this case, voice is a primary response to declining quality, and exit is only resorted to if voice fails as a remedy. This pattern of response requires that the product or service is originally superior in quality. If it is originally perceived to be very close in quality to its alternatives, then substitutability is high and exit is a stronger option. It also depends upon the perception that voice can be effective in restoring the original superiority in quality.
Exercising voice also takes relatively greater effort than exit. For this reason, voice is not as likely to be exercised by consumers of a large number of different products -- the cumulative cost becomes prohibitive. It is more likely to be used by members of organizations, because the number of groups people belong to is not generally too large, or for bigger purchases, where the cost of using voice could be justified by the expected gain.
The efficacy of voice is also limited in competitive markets with many actors. This makes voice more likely to exercised by members of organizations or participants in markets with few significant participants -- venues where the expectation of effective influence is greater.
Recognition of voice as an available corrective mechanism can prompt efforts to create institutions and mechanisms to make its user easier. This shows an important difference between exit and voice: exit requires nothing more than the availability of a different choice, while the exercise of voice requires creativity. (Hirschman calls it an art.) This creates a bias in favor of exit, because decisions about which option to use are based on past experience. But since the use of voice needs to be adapted for each circumstance, there will likely be exactly fitting past exercise to draw upon.