Age Dependence in Failure Rates of Firms: The Liability of Newness

Stinchcombe (1965, pp. 148-150) claimed that organizations face a liability of newness: new organizations fail at higher rates than old ones. His argu- ments supporting this claim pertain to both internal organizational matters and environmental relations. New organizations are vulnerable because their participants are strangers. Efficient organization requires trust among members; and trust takes time to build. New organizations are also vulner- able because they have to create organizational roles and routines. Inventing and refining roles and routines take time and effort precisely when organizational resources are stretched to the limit.

These considerations direct attention to the effects of environments be- cause new organizations, particularly those with a new form, must socialize members into new roles and routines. They cannot rely on generalized skills and experience available in the labor market. New organizations also have to develop relationships with existing organizations on which they rely for resources. So trust, which as we noted takes time to build, is important in organization-environment relationships too.

In Chapter 4 we derived the liability of newness from fundamental prop- ositions about the nature of selection processes in the modern world of organizations. We argued that social selection processes favor organizations and organizational forms that have high levels of reliability of performance and accountability, and that reliability and accountability in turn depend on a capacity to reproduce structure with high fidelity. By all accounts, new organizations have lower fidelity in reproducing structure from day to day. That is, fidelity of reproduction increases with age. So reliability and accountability increase with age; and mortality rates decrease with age. Empirical indications of a liability of newness provide indirect support for this line of argument.

In recent years evidence has accumulated in support of Stinchcombe’s hypothesis. In the broadest study, Carroll (1983) examined the liability of newness in 52 populations of firms in such diverse industries as retail sales, printing, chemical manufacturing, metal industries, and saloons. He used Gompertz and Makeham models to estimate the rate of decline in mortality rates with age and found that the rates do tend to drop sharply with aging. Freeman, Carroll, and Hannan (1983) explored these issues using richer data on newspaper firms, labor unions, and semiconductor manufacturing firms. They also used Gompertz and Makeham models and found that mortality rates drop sharply with age even when the initial size of organizations is taken into account.

The study by Freeman, Carroll, and Hannan (1983), hereafter cited as the FCH study, used preliminary and partial data on unions and semiconductor firms. They analyzed the histories of 476 national unions, the set of members of the AFL, the CIO, and the AFF/CIO along with some “major independent unions.” Here we update and replicate these analyses using somewhat more complete data, not restricted to the longer-lived, more successful members of the population. We also use different parametric models. We devote considerable attention to contrasting alternative representations of age dependence in failure rates because, as we noted earlier, the effects of aging are so strong that it is difficult to obtain meaningful estimates of other processes affecting mortality unless age dependence has been well represented.

One important issue has not been addressed in previous research on the liability of newness. This is the plausible possibility that age variation in mortality rates is “spurious” in the sense that it reflects only the operation of unobserved but age-independent heterogeneity in cohorts of organizations. Suppose that each cohort of organizations differs in terms of diffi- cult-to- observe characteristics such as managerial competence, political harmony among departments, and so forth. Assume further, contrary to Stinchcombe’s thesis, that these characteristics do not change with aging. Then those organizations with the characteristics that make them ill-suited to surviving in the socioeconomic environment will tend to experience mortality at early ages. As these events of mortality occur, the average mortality rate in the population will fall as a result of the loss of organizations with very high mortality rates. Since this process occurs on a time scale measured in age, the result of the unobserved heterogeneity is to give the empirical indications of age dependence in the rate. This is an instance of the general rule that unobserved heterogeneity and time dependence have similar empirical indications. The best way to combat this problem is to measure the relevant factors and include them as observed covariates, and we have indeed tried to do so. But it is also useful to examine the consequence of estimating models that allow the rate to depend on unobserved heterogeneity as well in order to see whether apparent age dependence is eliminated. We do so in this chapter using a parametric representation of the unobserved heterogeneity.

Source: Hannan Michael T., Freeman John (1993), Organizational Ecology, Harvard University Press; Reprint edition.

Leave a Reply

Your email address will not be published. Required fields are marked *