Data (updated using all available data as of 06/2020)
The following downloadable file provides firm-year and firm-quarter measures of earnings fidelity and economic persistence, for all firms with available data using over the period of 1980-2019 fiscal years. For a detailed definition of the variables, please see the paper below. Please let us know if you have any other questions.
Earnings fidelity data (annual frequency):
Earnings fidelity data (quarterly frequency):
Programs
The programs can be downloaded below.
[HMM_EarningFidelity] [Sim_EarningFidelity] [Sim_Example] [all programs and readme]
The main program “HMM_EarningFidelity” runs the Markov Chain Monte Carlo(MCMC) algorithm to estimate the main model parameters. To use the the main program, user needs to install R software and two Rcpp related packages: Rcpp and RcppArmadillo. Rcpp is an interface between R and C++. For more information on Rcpp files, see this page.
Once the Rcpp packages are correctly installed and the main program is loaded, the user needs to provide the following input variables to the program:
#Y: Observed earnings signal (0 or 1). A vector of size T=sum(T_i), where T_i is the number of observed earnings signal for firm i.
#Z: Observed covariates related to state transition. A matrix of size T by P where P is the number of covariates related to state transition.
#X: Observed covariates related to emission probability. A matrix of size T by L where L is the number of covariates related to emission probabilities.
#ID: Observed firm ID (from 1 to n). A vector of size T.
#a_sigma: standard deviation of the proposal distribution to sample firm-level intercept in the state transition function.
#b_sigma: standard deviation of the proposal distribution to sample firm-level intercept in the emission function.
#NB: number of burn-in MCMC draws.
#N: number of MCMC draws for parameter estimation.
The program provides the following ouput:
1) Approximate draws of parameters from the posterior distribution. These draws can be used to plot traceplot of parameters to check convergence and infer the main parameter values.
2) Estimates of firm-level intercepts in the state transition and emission function.
We recommend first-time users to first run the program on a simulated data generated by the code “Sim_EarningsFidelity.” “SimExample” provides instructions on how to estimate model parameters using the main program.