The 2019 Canadian Stata Conference was held on 30 May at the Banff Centre for Arts and Creativity.
9:20–9:40 | Session I: Causal inference
Abstract:
This presentation explores the impact financial considerations
have on firm export, investment behaviour, and productivity
growth using marginal treatment effects estimation. The research
has two goals. First, we attempt to quantify the positive
selection on unobserved heterogeneity in a firm's return to
exporting. This return may potentially reflect differences in firms'
access to credit markets. The first goal is to update the
learning-by-exporting literature to account for ex-ante differences
in financial health and credit constraints. Quantifying this difference
allows for improved understanding of the interaction between trade,
financial markets, and productivity growth. Second, this presentation also
provides a general framework under which we consider the minimal
conditions required for the identification of "constrained"
agents (firms) relative to their unconstrained counterparts. The
researcher observes only an agent's actual action, but not any
particular agent's constraints. We are likely to confound estimates
because we are unable to distinguish the reason an agent does not engage
in a behaviour, in this case, a firm not exporting or investing. There
is potential to confuse firms who expect low returns from an investment
and thus do not invest with those who could not invest because of credit
constraints. To overcome this identification problem for this set of
heterogeneous firms, we perform a marginal treatment effects estimation.
Specifically, we use the Stata module mtefe (Andresen 2018) to
assess exporting and a firm's productivity growth while accounting
for the potential of credit constraints.
Additional information: canada19_Petrunia.pdf
Robert Petrunia
Lakehead University
|
9:40–10:45 | StataCorp presentation
Abstract:
The increasing availability of high-dimensional data and increasing
interest in more realistic functional forms have sparked a renewed
interest in automated methods for selecting the covariates to include
in a model. I discuss the promises and perils of model selection and
pay special attention to some new estimators that provide reliable
inference after model selection.
Additional information: canada19_Drukker.pdf
David Drukker
StataCorp
|
11:00–12:10 | Session II: Household finance
Abstract:
Consumers face a choice when evaluating financial contracts: study the
fine print and incur a cognitive cost, or ignore it and risk costly
surprises in the future. We use a pair of policy changes in Chile to
contrast two measures to protect consumers from fine print; the first
improves disclosure, and the second standardizes and regulates contract
features. With administrative data from the banking regulator on
consumer loans, we use a regression discontinuity design to estimate the
causal effects of these regimes. Consumers offered standardized contracts
experienced 40% (14.4 percentage points) less delinquency. Using a
difference-in-differences design, we find that sophisticated borrowers
are helped most by increased disclosure, while unsophisticated borrowers
benefit more from product standardization. Additionally, we show that
only sophisticated borrowers who benefit from the informational
disclosure treatment leave less "money on the table." We contextualize
these results in a stylized model that predicts that financially
sophisticated borrowers will benefit from disclosure, while
unsophisticated borrowers will benefit from standardization based on
differentials in the cost of studying.
Additional information: canada19_Kulkarni.pdf
Sheisha Kulkarni
University of Virginia
Abstract:
Using Canadian consumer microcredit data from 2010–2018,
I analyze the relationships between consumers' home
equity line of credit (heloc) utilization, macroeconomic
factors, and consumer-specific characteristics. Major
shares of heloc loans are held by borrowers with high credit
quality at origination. Our estimates show that a 10 percent
increase in local house price leads to about a $1,600 increase
in outstanding heloc loan amount. However, the heloc loans
as a ratio of local house prices do not respond to house
prices in terms of economic significance, implying that consumers
may have targeted their loans to certain leverage ratio. Neither
the heloc interest rate nor the provincial unemployment rate has
an economic significant relationship with heloc outstanding loan
amounts. Regarding heloc delinquency, we found that a majority
of accounts in arrears exhibit over 80 percent utilization.
Delinquency is mainly driven by consumer-specific factors, and
macroeconomic factors except house price changes have only minor influences.
Anson Ho
Bank of Canada
Abstract:
In this presentation, I am investigating the following "puzzle":
in aggregate share trends, cash seems displaced by contactless
credit card (CTC) payments; however, at the micro level, regression
analysis finds no significant effect of CTC on cash once unobserved
heterogeneity is accounted for. I relax the assumption of homogeneous
coefficients across all individuals and evidence the existence of
different micro substitution effects that are confounded in aggregate
data. All the data analysis is conducted in Stata using, i.a., panel
data analysis, cluster analysis, extended linear regression and finite
mixture models.
Additional information: canada19_Felt.pdf
Marie-Helene Felt
Bank of Canada
|
1:30–2:30 | StataCorp presentation
Abstract:
Dynamic stochastic general equilibrium (DSGE) models are used in
macroeconomics for policy analysis and forecasting. A DSGE model
is a system of
Additional information: canada19_Schenck.pdf
David Schenck
StataCorp
|
2:45–3:55 | Session III: Survival models
Abstract:
This presentation will discuss a new community-contributed command,
ipdmidas, that facilitates implementation of models for
meta-analysis of individual patient data (IPD) from both a
frequentist and Bayesian perspective. Meta-analysis of diagnostic
test studies typically involves synthesizing aggregate data (AD)Ben,
such as the 2 x 2 tables of diagnostic accuracy. Bivariate
random-effects meta-analysis (BREM) and the hierarchical summary
ROC (HSROC) model can appropriately synthesize these tables, leading
to clinical results such as the summary sensitivity and specificity
and summary ROC(SROC) curves across studies. However, translating
such results into practice is often limited by between-study
heterogeneity and application to some "average" patient across
studies. Meta-analysis of IPD examines study-level covariates,
explains the between-study heterogeneity, and examines patient-level covariates,
assessing the effect of patient characteristics on test accuracy.
This allows tailoring of test results to the individual patient and
informs individual diagnostic strategies. I will show how BRMA/HSROC
models have been and maybe extended to IPD to depict how covariates
affect sensitivity, specificity, and between-study heterogeneity and
correlation. I will also demonstrate how ipdmidas can be used to obtain
such metrics and other informative results such as the diagnostic
odds ratio, the positive and negative likelihood ratio tests, and the SROC curve.
Additional information: canada19_Dwamena.pdf
Ben Adarkwa Dwamena
University of Michigan Medical School
Abstract:
This presentation proposes an empirical approach for jointly modeling the
impact of the minimum wage on the wage distribution and on movements
in and out of the workforce. We estimate the effects of the minimum
wage on the hazard rate for wages, which provides a convenient way
of rescaling the wage distribution in the presence of employment
effects linked to the minimum wage. We use the estimates to decompose
the distributional effects of minimum wages into effects for workers
moving out of employment, workers moving into employment, and workers
continuing in employment. The estimator is implemented in Stata using
the command for generalized linear models.
Additional information: canada19_Townsend.pdf
James Townsend
University of Winnipeg
Paper abstract:
The launch of new business ventures is an important source of dynamism
for both advanced and transitioning economies. However, survival
prospects are low and many new business ventures remain small. Yet,
the empirical evidence from administrative level data suggests that
much of aggregate employment and productivity gains stem from a small
subset of successful,including high-growth, startups; however, these
data often lack information on firm strategy,financing, innovation
activities and founder characteristics, among other variables. Using a
novel detailed survey dataset, the Kauffman Firm Survey, we study a
representative cohort of American startup firms launched in 2004 over
an eight-year period until 2011; overlapping with the business cycle
pre and post the Great Recession of 2008-2009. Considering a rich set
of firm-level factors including financing conditions, we examine the
role of innovation measured by the firms industrial technology sector
as well as self-reported innovation and R&D activities in driving firm
survival and performance. We also investigate the role of innovation
in securing external financing as a potential mechanism in early stage
firm growth.
Presentation abstract: This presentation proposes an empirical approach for jointly modeling the impact of minimum wage on the wage distribution and on movements in and out of the workforce. We estimate the effects of the minimum wage on the hazard rate for wages, which provides a convenient way of rescaling the wage distribution in the presence of employment effects linked to the minimum wage. We use the estimates to decompose the distributional effects of minimum wages into effects for workers moving out of employment, workers moving into employment, and workers continuing in employment. The estimator is implemented in Stata using the command for generalized linear models. Leonard Sabetti
Universite d'Auvergne Clermont-Ferrand I
|
4:10–4:40 | Session IV: Pedagogy
Abstract:
Can instructional videos help improve learning in undergraduate
courses in business statistics? With the proliferation of online
learning materials available as MOOCs, institutions of higher
learning are also actively experimenting with innovative models
of blended learning. Online content presents an interesting
opportunity for instructors to complement their face-to-face
interactions with students with online videos and other digital
resources. I present the results of an experiment with a large
number of students (n = 1202) who were enrolled in the second course
in business statistics in a business faculty at a North American
University. The students, without their knowledge, were randomly
divided into control and treated groups. Students in the treated
group were encouraged (not required) to watch brief videos that
reinforced concepts already introduced in the lecture hall. Students
in the control group were not advised as such. The treated group was
monitored online to measure individual students' level of interaction
with the online content (especially videos). Afterwards, students'
performance in assignments and exams was compared between the treated
and control groups while controlling for the level of engagement with
the online materials. Furthermore, students' responses to questions
with the highest correct and wrong responses were analyzed. This
presentation describes the findings from the experiment and shares
insights for reinforced learning using online videos to improve
learning and pedagogy in undergraduate courses in business statistics.
Murtaza Haider
Ryerson University
|
4:40–5:15 | Open panel discussion with Stata developers |
Presentations from StataCorp
XX
XX
XX
XX
Scientific committee
Matthew Webb (Chair)
Carleton University
Kim Huynh
Bank of Canada
Leslie-Anne Keown
Carleton University
Matthias Schonlau
University of Waterloo
Vicki Stagg
Calgary Statistical Support
Presenters from StataCorp
David Drukker
David M. Drukker is the Executive Director of Econometrics at Stata. His passion for programming enabled him to start working at Stata in 2000, when he finished his Ph.D. in economics at the University of Texas at Austin. Since then he has developed many Stata commands for estimating treatment effects and for analyzing panel data, time-series data, and cross-sectional data. He played a key role in the initial development of Stata MP, helped integrate Mata into Stata, and helped develop Stata and numerical techniques.
David Schenck
David Schenck is a Senior Econometrician at StataCorp and is the primary developer of Stata's new DSGE features. He has a bachelor's degree in economics from Vanderbilt University and a master's degree in economics from Boston College. His research interests lie in macroeconometrics.