Family firms are the most prevalent firm type in the world, particularly in emerging economies. Dynastic family firms tend to have lower productivity, though what explains their underperformance is still an open question. We collect new data on CEO successions for over 800 firms in Latin America and Europe to document their corporate governance choices and, crucially, provide causal evidence on the effect of dynastic CEO successions on the adoption of managerial best practices tied to improved productivity. Specifically, we establish two key results. First, there is a preference for male heirs: when the founding CEO steps down they are 30pp more likely to keep control within the family when they have a son. Second, instrumenting with the gender of the founder’s children, we estimate dynastic CEO successions lead to 0.8 standard deviations lower adoption of managerial best practices, suggesting an implied productivity decrease of 5 to 10%. To guide our discussion on mechanisms, we build a model with two types of CEOs (family and professional) who decide whether to invest in better management practices. Family CEOs cannot credibly commit to firing employees without incurring reputation costs. This induces lower worker effort and reduces the returns to investing in better management. We find empirical evidence that, controlling for lower skill levels of managers, reputational costs constrain investment in better management.
JOB MARKET PAPER
International Data on Measuring Management Practices, 2016. (with N Bloom, R Lemos, R Sadun and J Van Reenen)
American Economic Association – Papers & Proceedings
Rapid advances in computer power and increased openness of national statistical agencies have led to unprecedented availability of large datasets. Consider three types of firm datasets. First, governments collect administrative data on firms: information on jobs, investment and output has long been collected to calculate national, industrial and regional statistics. Second, there has been an explosion of Big Data – various forms of data typically created for business purposes. Products like ORBIS contain over 50 million firms from almost every country in the world and can be used to address many questions. Another example is Compustat, which contains extensive data for about 6,000 listed US firms but excludes the other 99% of private firms. We focus on a third type of international firm data, which is collected from surveys. In an age of rich administrative and Big Data why bother with such surveys?
Over the last decade the World Management Survey (WMS) has collected firm-level management practices data across multiple sectors and countries. We developed the survey to try to explain the large and persistent total factor productivity (TFP) differences across firms and countries. This review paper discusses what has been learned empirically and theoretically from the WMS and other recent work on management practices. Our preliminary results suggest that about a quarter of cross-country and within-country TFP gaps can be accounted for by management practices. Management seems to matter both qualitatively and quantitatively for performance at the level of the firm and the nation. Competition, governance, human capital, and informational frictions help account for the variation in management. We make some suggestions for both policy and future research.
PAPERS IN PROGRESS
Recent research has documented growing wage disparities in between-firm wages. One possible reason is that the heterogeneous application of modern management practices has led to increasing firm-specific productivity differences and induced greater worker sorting. We use matched employer-employee data from Brazil linked to the manufacturing firms covered in the World Management Survey (WMS) to characterize different personnel management profiles firms use to affect recruiting, motivation, and retention of high-quality workers. We show that the heterogeneity in pay across firms has two key components that have been conflated in prior work: a level component and a tenure component, both of which can vary across occupations, and drift over time. Because they are associated with management practices aimed at recruiting and retaining talented workers, they also help explain observed increases in sorting between high-wage workers and high-wage firms.
This paper uses two new datasets from India to study management and productivity in public and private schools. We report four main sets of results. First, we conduct a systematic measurement of management practices in public and private primary schools and find substantially higher management scores in private schools. These gains are mainly driven by differences in people management as opposed to operations management. Second, we find that the private school advantage over public schools in student value-added is largely accounted for by differences in management practices. Third, we find that the private school advantage in measures of people management is consistent with independent measures of personnel policy. Specifically, private school teacher pay is positively correlated with measures of teacher value-added, and private schools are more likely to retain teachers with higher value-addition and let go teachers with lower value-addition. Neither pattern is seen in public schools. Finally, measures of school-management quality are more correlated with measures of effective teacher practice in public schools than in private schools. These results suggest that once a school can optimize their selection and retention practices – that is, get their personnel policy right – continuous oversight of what happens in the classroom is less important. When this channel is shut down (as is the case in public schools), school management may be more important.
Where the money goes: school management and resource allocation in Brazil
with D. Clarke and T. Teodorovicz
Teachers account for one of the largest chunks of expenditures of public school systems, and teachers are also one of the key inputs into the education production function. Public school systems are often constrained in wage-setting and cannot actively implement incentive-pay schemes. So what works in improving teacher performance in public schools? We explore a change in the funding formula for public schools in Brazil that mandates an exogenously set share of resources that must be used in the teacher wage bill of school systems. This change affected Brazil’s municipalities differently as some received different amounts of funds, and chose to spend their funds in different ways. In this paper, we first provide a long-term look at 20 years of data and document how teacher salaries, number of teachers employed and student-teacher ratios have changed over time in Brazilian municipalities. Secondly, we employ an event study to document how different municipalities responded to the funding change and explore the characteristics that may help explain the heterogeneity in resource usage as well as student outcomes. We propose that management practices across different school systems explains part of the superior/inferior ability of school systems to effectively use and distribute the increase in resources they experienced after the funding formula change.
SELECTED ONGOING PROJECTS
A large share of firms in developing countries are badly managed and in Brazil, the worst managed firms are in the North and Northeast regions. Most Northeastern firms are family owned and the evidence suggests that these firms adopt fewer good management practices. This project aims to understand what are the key barriers for adoption of best management practices. Firms will be stratified by ownership status so that the interventions can be tested among family and non-family firms. The planned RCT will use a stratified randomized control trial to test (1) whether giving firms information on their ranking relative to peers in terms of their management practices along with advice on how address the inefficiencies affects the adoption of such practices and performance indicators, and (2) whether offering a highly-subsidized consulting service program from a Brazilian non-profit consultancy (SENAI) results in better management practices and performance indicators. To test whether take-up of the consulting services is hindered by credit constraints, a share of the managers who get the information-only treatment will also be offered a line of credit with the Northeast Bank of Development to hire these consulting services. Our design includes a rich data collection exercise, including measuring the risk preferences and cognitive abilities of founder and family CEOs to consider how these factors may relate to adoption of new technologies. We will also be able to merge this data with employer-employee data measuring worker flows, with productivity data from the annual census of manufacturers, as well as credit records from the Northeastern Bank of Development. This is a large project that aims to answer the following research questions: A) do they not know they are bad? (information treatment) B) are they credit constrained? C) do firms under/over-estimate the returns to consulting services? C) are family CEOs too risk averse? (risk preferences measurement)
Family firms, female leadership and management in Denmark
with M. Bennedsen
We have set up a self-reported survey to collect data on management quality of firms in Denmark, mirroring the US Census Management and Organizational Practices Survey. We have also included questions aimed at understanding basic the time use of managers (including categories such as meetings, coaching, mentoring, family, etc.). Our expected sample size is 15,000 firms in Denmark, and will be matched to some of the richest datasets available and a number of interesting research questions become viable. One key question to explore circles around issues of gender in management styles and productivity. The research question is: do female CEOs have different management styles and what impact does this have on productivity and workers?
Developing management: an expanded evaluation tool for developing countries (RISE Working Paper 7) [project website] [slides]
with R. Lemos
Coverage: World Bank blog
There is striking evidence showing a large tail of badly managed schools and hospitals in developing countries across several management areas such as operations management, performance monitoring, target setting and people management. But where exactly, along the process of setting their management structures, are these organizations failing? This paper presents new evidence from an expanded survey tool based on the World Management Survey instrument. We collect detailed data using face-to-face interviews in settings where weak management practices prevail and observe more variation in the left tail of the distribution. Using this data, we explore three main “activities” within each management area: (1) process implementation, (2) process usage, (3) process monitoring efficiency and frequency. We have collected data with schools in India and Mexico and are working with teams surveying schools in Colombia and hospitals in China and India.
Governments of 7 states in Mexico are implementing a management upgrading program for the public schools in their states. The programs include training on classroom observations, training on tools for measuring writing, reading, and math skills, and training on management and leadership practices for school directors. States have carried out the implementation of this intervention to varying degrees and using different delivery approaches. In this project, we seek to understand “what works” in improving management in public schools, and the effect different approaches may have in take-up and student outcomes. We measured the quality of management in Mexican schools in these states using the Development WMS tool at baseline (2014-15), at the first follow up (2016-2017) and have one more wave scheduled for 2018. We have also collected data on the fidelity of the intervention take-up (school director training), and of school director movement across schools. At the end of the project we will also merge the data with student outcomes from the schools in the D-WMS sample.