JBNST - Vol. 238/2-3 - 2018

This paper uses business tax administrative data to describe the annual firm growth rate distribution in Canada over the 2000–2009 period. This administrative tax database provides a unique lense to study firm growth as it allows us to look at the universe of Canadian employer firms and investigate the firm growth distribution across different dimensions. A non-normal, fat-tailed shape for the firm growth distributions holds across years, industries, regions, as well as firm size and age classes. The results show that the distributions of employment growth rates in Canada have more density in both the center and tails than a normal distribution. The evidence paints a picture of firm growth dynamics whereby most firms change very little each year, while a nontrivial amount also markedly grow or decline. A final finding is that young firms, aged four or less, represent a special case with an upwardly skewed distribution and a median growth rate greater than zero.

Most scientists consider randomized experiments to be the best method available to establish causality. On the Internet, during the past twenty-five years, randomized experiments have become common, often referred to as A/B testing. For practical reasons, much A/B testing does not use pseudo-random number generators to implement randomization. Instead, hash functions are used to transform the distribution of identifiers of experimental units into a uniform distribution. Using two large, industry data sets, I demonstrate that the success of hash-based quasi-randomization strategies depends greatly on the hash function used: MD5 yielded good results, while SHA512 yielded less impressive ones.

We investigate estimation and inference of the (local) average treatment effect parameter when a binary instrumental variable is generated by a randomized or conditionally randomized experiment. Under i.i.d. sampling, we show that adding covariates and their interactions with the instrument will weakly improve estimation precision of the (local) average treatment effect, but the robust OLS (2SLS) standard errors will no longer be valid. We provide an analytic correction that is easy to implement and demonstrate through Monte Carlo simulations and an empirical application the interacted estimator’s efficiency gains over the unadjusted estimator and the uninteracted covariate adjusted estimator. We also generalize our results to covariate adaptive randomization where the treatment assignment is not i.i.d., thus extending the recent contributions of Bugni, F., I.A. Canay, A.M. Shaikh (2017a), Inference Under Covariate-Adaptive Randomization. Working Paper and Bugni, F., I.A. Canay, A.M. Shaikh (2017b), Inference Under Covariate-Adaptive Randomization with Multiple Treatments. Working Paper to allow for the case of non-compliance.

We analyze a data set containing rental revenues, maintenance costs, and sale prices of five different types of rental machines to econometrically estimate key relationships needed to implement a dynamic programming model of the optimal timing of replacement of rental equipment owned by a large multilocation firm in the equipment rental industry. The model reveals significant potential to improve rental company profitability by improving the strategic timing of equipment replacement. The gains from the optimal replacement strategy come from exploiting seasonal variation in rental demand and the timing of the business cycle due to their effects on rental revenues and the cost of replacement. For some machines we find the optimal replacement strategy is procyclical, but for others we find that a countercyclical replacement strategy — where replacements are concentrated in slow periods of the business cycle — can significantly increase firm profits.

In this paper, we study a credit risk (collateral) management scheme for the Canadian retail payment system designed to cover the exposure of a defaulting member. We estimate ex ante the size of a collateral pool large enough to cover exposure for a historical worst-case default scenario. The parameters of the distribution of the maxima are estimated using two main statistical approaches based on extreme value models: Block-Maxima for different window lengths (daily, weekly and monthly) and Peak-over-Threshold. Our statistical model implies that the largest daily net debit position across participants exceeds roughly $1.5 billion once a year. Despite relying on extreme-value theory, the out of sample forecasts may still underestimate an actual exposure given the absence of observed data on defaults and financial stress in Canada. Our results are informative for optimal collateral management and system design of pre-funded retail-payment schemes.