next up previous
Next: Hospitalization and Mortality Up: Latent Variable Models for Previous: New Methods for

Some econometric duration models

  1. Sequence of probits: failure when

    1. are s and duration:
    2. are 's
    3. Hazard is
    4. is linear Normal regression;
    5. are independent truncated Normals.
    6. McCulloch and Rossi(1994).

  2. Heckman(1981) gives many generalizations of this discrete time latent variable structure.

  3. Continuous time probit sequence. Trials in a Poisson process of rate .
    1. Hazard is
    2. are 's and numbers of trials.
    3. is linear Normal regression.
    4. is Gamma.
    5. are independent truncated Normals and Poissons.

  4. Introduce fixed effects.

    1. Hazard is , given .
    2. Simplify functional form. Hazard is
    3. Lancaster(1990)

  5. Lancaster(1972) has failure when

    Latent variable is a Weiner process; this is an autocorrelated variant on the probit sequence. Jovanovic(1984) embeds this in a model of job tenure. Failure is job separation. Component conditionals seem straightforward but have not been done.

  6. Structural search models. Neumann, Ridder, Kiefer, Lancaster et al.
  7. For overview, Ridder(1995)
    1. are duration of search,; accepted wage offer.
    2. are number and value of rejected offers.



Tony Lancaster

tl@pstc3.pstc.brown.edu

Last modified: 12 September 1995