The management and organizational practices survey (MOPS): An overview 1

1. Introduction

The important role of management in the success of firms has long been stressed by academics in business and management, the media, and consultants, but most evidence has been anecdotal or based primarily on case studies. We describe one of the innovative steps forward in measuring management practices: the development and fielding of the first ever large-scale survey of management in the United States, the Management and Organizational Practices Survey (MOPS). The MOPS was developed as a partnership between the Census Bureau and an external research team of Nick Bloom (Stanford), Erik Brynjolfsson (MIT), and John Van Reenen (MIT), and later Steven Davis (University of Chicago) and Kristina McElheran (University of Toronto), and was sent to about 50,000 manufacturing establishments in 2011 and 2016. In this paper, we provide the background and motivation for developing the MOPS by describing the existing empirical literature on management practices, uncertainty, and data and decision making.

Already the MOPS has had wide-ranging impacts on the study of management practices worldwide, as questions based on the MOPS have been or will soon be issued as part of censuses in Canada, Germany, Pakistan, Japan, Australia, and the United Kingdom [1]. The statistical agencies of Pakistan and Mexico have issued surveys that were adapted from the MOPS, although these surveys were conducted face-to-face rather than with paper instruments or electronically due to the fact that mail and e-mail were considered unreliable for contacting firms [2].

Syverson [3] notes that academic writing on the importance of management for profitability dates back at least to Francis Walker [4], but it has only been in the post-war period that management has been considered explicitly in the study of firms. Early “managerial” models of the firm [5] focus on principal-agent problems, wherein a manager of a firm may seek to solve a different objective than her profit-maximizing employer. A small theoretical literature developed around a more robust model of role of management in firm structure starting in the early 1990’s, but meaningful empirical studies of the role of management began to supplement these early theories only much later.

The theoretical literature on management that developed starting with Radner [6] largely focused on incorporating the anecdotal evidence available in the business press and aggregate data into models of firm structure. Radner’s interest in management stems largely from the observation that the growing number of large firms must require a more complex internal structure than the simple model of a profit-maximizing agent, or even a principle-agent model, allows. While Radner’s motivations are not rooted in extensive empirical study of the role of management, this small literature has had far-reaching implications, including motivation for macroeconomic models of rational inattention [7]. Milgrom and Roberts propose a theoretical model of technological adoption that exhibits complementarities with changes in work practices and firm organization [8].

Recent findings on productivity have shown that there is significant and persistent dispersion of productivity across firms and even establishments that can only partially be explained by differences in inputs [9], production technologies, price heterogeneity [10], and idiosyncratic shocks [11]. Based on the pre-existing theoretical research and anecdotal evidence regarding the importance of management practices, the hypothesis was put forward that perhaps management practices could account for some of the firm- and establishment-level heterogeneity in productivity.

Unlike these studies of firm- and establishment-level heterogeneity in productivity, which were made possible by the availability of representative or even population-level microdata from government sources, empirical studies of management were virtually non-existent until ten years ago. Syverson goes so far as to state that “perhaps no potential driver of productivity differences has seen a higher ratio of speculation to actual empirical study” [3]. Several recent studies have begun to alter this ratio, however, by creative uses of existing datasets.

Bertrand and Schoar use publically available data to match CEOs to firm performance data and find that demographic data about the CEOs predict management style [12]. Ichniowski et al. [13] and Bartel et al. [14] examine the impact of changing management practices on productivity in industry-specific samples of steel finishing plants and valve manufacturing, respectively. Ichniowski et al. [13] and Bartel et al. [14] develop specific surveys of the human resource management practices for their respective samples; the latter also considers complementary IT investment. Acemoglu et al. use measures of decentralization from two French data sets (Changements Organisationnels et Informatisation and Enquête Response) and a British data set (Workplace Employee Relations Survey) as proxies for delegation of decision making to managers [15]. Related work by McElheran links the private Harte Hanks Computer Intelligence database to performance data from the 1997 Census of Manufactures to examine decentralization of decision making within multi-unit firms [16].

In addition, a sizeable literature in the field of development economics has taken shape over the past five years focusing on the business training of microenterprises. This literature is primarily experimental in nature, offering business training to selected entrepreneurs, with mixed results. See Karlan et al. [17] and McKenzie and Woodruff [18] for surveys of this literature. De Mel et al. also constructed a survey tool to gauge the existing management skills of entrepreneurs in Sri Lanka [19]. Similarly, Bloom et al. conducted a field experiment on 17 Indian textile firms having between 100 and 1,000 employees wherein the experimental firms were given management training, and performance was extensively monitored during and after the training period [20].

More ambitious direct measurement efforts have also taken shape. Several large-scale, multi-industry surveys were recently developed and administered. One of these, the World Management Survey (WMS), is of especial interest since it has served as a starting point of a sort for the MOPS. The WMS, started in 2004, has run extensive double-blind telephone interviews on management practices with over 11,300 organizations in manufacturing across 34 countries between 2004 and 2014, and its methodology has been extended to samples in the retail, education, and healthcare industries in that time. As detailed below, the WMS has been adapted by international organizations for a survey, and Statistics Canada has also developed two related surveys.

The paper proceeds as follows: Section 2 provides an overview of existing management surveys, Section 3 describes the core content of the MOPS, Section 4 discusses data and decision making, Section 5 discusses uncertainty, and Section 6 provides some discussion about future directions and concludes.

2. Existing management surveys and research

Management practices have long been used as an explanation for the residual firm- and establishment-level heterogeneity in productivity that could not be explained by other, more measurable factors, even in the absence of strong empirical support. However, increasingly economists and government agencies have conducted surveys in an effort to measure management practices. Tables 1 and 2 provide an overview of these surveys; we discuss each in turn below.

The most widely cited empirical study of management at this time, the WMS, uses 18-question telephone interviews to gather evidence regarding the importance of management practices. A summary of the practices of the WMS is offered in Bloom et al. [21] and a synopsis is given here. The WMS hires students in Master of Business Administration (MBA) or similar programs to call mid-level managers of firms in manufacturing, healthcare, education, and retail in 20 countries. Each interview is conducted in the native language of the interviewee, and the calls last 45 minutes on average. The interview questions are open-ended, and then the interviewers score the responses on a scale from one (worst) to five (best).

The interviewee is not aware that the responses are scored, nor is the interviewer provided information about the firm’s performance when conducting the interview; moreover, the sample firms are chosen so that the interviewer is unlikely to have prior knowledge of the firm. The firms’ performance and financial data are obtained from independent sources. The interviewees are randomly selected from the population of all medium-sized firms in the given industry and country; that is manufacturing and retail firms that have 50–500 employees, hospitals that deliver acute care, and schools that educate 15-year-old students.

The questions asked of the interviewee fall into three categories: monitoring, targeting, and incentives/personnel management. The questions on monitoring ask the extent to which firms measure performance within the firm and use that data (if collected) to improve performance. The questions on targeting attempt to gauge how well firms set forward-looking goals and course correct if those goals are not met. Incentives/personnel management questions examine how employees are promoted, rewarded, and retained, or alternately reprimanded and dismissed.

Bloom and Van Reenen present the first results of the WMS finding that greater implementation of “structured management practices” – that is, increased monitoring of firm activity, implementation of clear targeting practices, and the presence of strong incentives for achieving the establishment’s targets – is associated with higher productivity, profitability, and survival rates [22]. They also compare cross-country results and find that U.S. firms generally implement more structured management practices than European firms, although there remain high levels of within-country dispersion of practices. Poor management practices are frequently associated with weaker competitive pressures and firms practicing primogeniture.

Bloom et al. examine the management practices of multi-national firms and find that firms that exist across countries with high levels of trust tend to decentralize decision making [23]. Establishments of multinational firms tend to have high levels of structured management practice implementation in general. Bloom et al. find that private equity owned firms have more structured management practices than do government, family, or privately-owned firms, particularly in monitoring practices [24]. Private equity firms are also more likely to be structured in a way that grants more autonomy to individual establishments relative to other types of firms.

Bloom et al. [21] note that there are high levels of dispersion in adoption of structured management practices for schools [25] and hospitals [26], with government-run schools and hospitals generally having lower scores on structured management scores than their privately-owned counterparts. Other users of the WMS methodology have found a spectrum of adoption of structured management practices in fostering, adoption, and nursing homes [27]; tax agencies in OECD countries [28]; public-private partnerships [29]; substance abuse clinics [30]; UK university departments [31]; tradable service firms in Ireland [32]; Nigerian civil service [33]; and American hospitals and cardiac care units [34, 35]. Rasul and Rogger also find that ethnic diversity at public sector organizations in Nigeria is positively correlated with structured management practices [36]. Rahaman and Al Zaman [37] use the Bloom and Van Reenen [22] WMS data set with Loan Pricing Corporation DealScan data to show that structured management practices are negatively correlated with interest rates on corporate loans and that firms with more structured practices are more likely to borrow from more reputable lenders than firms with less structured practices.

In 2008 and 2009, the European Bank for Reconstruction and Development and the World Bank adapted the WMS to conduct the Management, Organisation, and Innovation survey (MOI) to study management practices in 10 transition countries. Although the 12 questions on the MOI survey instrument were adapted from the WMS, the questions were closed rather than open-ended, and interviews were conducted face-to-face rather than over the telephone. Using MOI data, Bloom et al. find that management scores in Central European transition countries are quite similar to management scores in Western Europe, while Central Asian transition countries trail other developing Asian countries in structured management practice adoption [38].

The National Employer Survey (NES), conducted by the Census Bureau over three waves (1993, 1997, and 2000), asked questions related to employees and employment, employee training, business characteristics, and equipment and technology. The NES had 3,358 respondents for 1993 and 5,465 respondents for 1997 (and a longitudinal component). Supplements on partnerships between employers and schools were conducted by telephone interview in 1996 and 1998. A third wave of the NES was run in 2000, sampling 2,825 establishments that responded to the second wave of the survey as well as 50 employees each for 225 matched establishments. The establishment component of the NES, which was a joint venture with the National Center for the Educational Quality of the Workforce, was conducted as a computer-aided telephone interview of plant managers. Cappelli [39] provides a detailed overview of the NES.

Cappelli and Neumark use NES data and find weak evidence of a positive impact of increased decision making power for employees on productivity [40]. Black and Lynch find that unionized establishments with increased worker decision making have higher productivity than do similar nonunion establishments and unionized firms with traditional decision making structures [41]. Establishments with higher education levels are more productive than establishments with lower education levels, and establishments with more computer usage by non-managers are more productive than establishments that where non-managers are less likely to use computers.

Statistics Canada conducted the Workplace and Employee Survey (WES) annually on a representative sample of approximately 10,000 to 15,000 establishments between 1999 and 2006 that included questions on compensation, training, human resources practices, organizational change, performance, business strategy, innovation, and technology use. Statistics Canada also ran the Survey of Innovation and Business Strategy on roughly 4,000 and 8,000 establishments in 2010 and 2013, respectively. The establishments were drawn from fourteen industries as defined by the North American Industry Classification System (NAICS). The survey sought to gather information on monitoring, structure, use of advanced technology, human resource management, and other business strategies.

Statistics Canada’s WES is conducted in two parts: a computer-aided phone survey is administered to employers and a telephone interview conducted with employee participants. The survey covered a longitudinal sample of establishments, with approximately 9,000 establishments selected in 1999, with new establishments gradually added (and naturally other establishments exiting), leading to a sample of approximately 15,000 units in 2005. The establishments are selected to be representative of workplaces in Canada. The employer survey consists of 50 questions divided into nine sections: workforce characteristics and job organization, compensation, training, human resources practices, collective bargaining, workplace performance, business strategy, innovation, and technology use.

The employee sample consists of no more than 24 randomly-selected employees per establishment, with an annual sample of about 20,000 workers. Employees are surveyed for two years, and then a new sample is drawn. The employee survey consists of 59 questions across ten categories: job characteristics, computers and other technologies, training and development, career-related training, employee participation, personal and family support programs, worker representation and industrial relations, compensation, work history/turnover, and demographics.

Yang et al. use the employer component of the WES to show that adoption of structured management practices is strongly correlated with particular business strategies of for-profit firms [42]. These strategies are: novelty, low-cost, and high-quality. Firms pursuing “novelty” strategies seek to provide a good or service that is unique in itself. Firms pursuing low-cost or high-quality strategies seek to compete on either cost or quality. Low-cost firms tend to delegate more to managers, whereas novelty firms tend to implement more autonomy for all workers. Both high-quality and novelty firms are likely to implement structured management practices related to incentives. Hong et al. also use the employer component of the WES to show that performance-based pay systems are complementary to decentralization of decision making from principals to managers [43].

The Survey of Innovation and Business Strategy (SIBS), also from Statistics Canada, takes representative samples of approximately 4,000 and 8,000 establishments in 14 NAICS industries in 2010 and 2012, respectively. The survey consisted of over 100 questions on business strategies and monitoring, enterprise structure, operational activities, relocation of activities in to and out of Canada, sales, relationships with suppliers, technology usage, innovation, structured management practices, and use of government support programs. This survey was sent to establishments both as a paper and an electronic survey form.

Brouillette and Ershov [44] use the SIBS to construct a measure of management practices that is analogous to the index created by Bloom and Van Reenen [22] and find that larger firms implement more structured practices. They find that this measure is positively correlated with a measure of business innovation for all sectors, but only in manufacturing industries are structured practices positively and significantly correlated with sales and profits.

3. Management and organizational practices survey

The Management and Organizational Practices Survey (MOPS) collects information on targeting, monitoring and incentives managerial practices; the locus of decision making within the organizational structure of the firm to which the establishment belongs; and related background information from a sample of U.S. manufacturing establishments. The 2010 survey consisted of 37 questions in three sections: management practices, organization, and background characteristics. The 2015 survey consists of 47 questions covering the original (modified) sections and new sections on data and decision making and uncertainty. In this section, we discuss the overall sample and collection strategies and the three common sections. Sections 4 and 5 discuss the new sections. The online appendices contain the complete instruments for 2010 and 2015, respectively.

4. Data and decision making

We start by providing motivation for the MOPS questions on data and decision making (two in 2010 and six in 2015) by reviewing the existing literature and research in this field. Part of the impetus for including management in theoretical economic models such as Radner [6] or Adam [7] is that managers may be essential for gathering and processing information. Indeed, two of the components of the structured management practices above, monitoring and targeting, can be described as a form of information processing. Management gathers information about production conditions both within and outside of the establishment (or firm) and then uses that information to set targets and make adjustments to the production process. The degree of data collection performed by firms is a key component of this relationship.

The rise of information technologies (IT) has made it possible for establishments to utilize ever increasing amounts of data in their decision making, and Brynjolfsson and Mendelson argue that the increasing availability of data has necessitated the development and implementation of structured management practices [54]. Much of the existing work in this field is focused on the implementation of information technologies. While IT and data and decision making (DDD) are clearly complementary, they are not necessarily proxies for one another. A firm could conceivably gather data for decision making without high levels of IT investment, while a firm that utilizes modern IT may not necessarily fully integrate DDD.

Bresnahan et al. use a combination of a telephone survey of 379 firms, computer capital data from Computer Intelligence InfoCorp, and input and output data from Compustat [51]. The telephone survey included 14 questions related to the organization of the firm’s workforce, which are neither fully orthogonal to nor entirely consistent with the definition of structured management practices given above. The survey measures uses of teams, dispersion of authority, and education. A detailed description of the data set is available in Brynjolfsson and Hitt [55].

Bresnahan et al. find that IT implementation and workplace reorganizations focused on teamwork and individual authority are both positively correlated with productivity and have complementary effects when implemented together [51]. Similarly, Aral et al. find high levels of complementarities between IT implementation, implementation of performance pay, and human resource management practices that monitor performance and give employee feedback [52]. Taken together, these three practices have a large positive impact on worker productivity in the 189 firms surveyed by a non-profit organization that educates firms on human resource practices that also purchased an IT system called Human Capital Management.

Bloom et al. combine the WMS with a private software utilization data source called Harte-Hanks [56]. They find that increased implementation of information technology leads to more decentralization in manufacturing firms, while implementation of communication technology leads to greater centralization.

The Census Bureau collected the Computer Network Use Supplement (CNUS) to the ASM sample in 1999. Like the MOPS, this data could be readily matched to high quality performance data from the ASM. Atrostic and Nguyen use this data to find that establishments that have computer networks have higher labor productivity than establishments that do not have computer networks [57]. They find that moving from the 10 th to the 90 th percentile of computer network use was associated with a 7.2% increase in labor productivity, as well as evidence that establishments with low labor productivity in earlier periods use the introduction of computer networks to “catch up” with more productive establishments. Additionally, the use of networks in 1999 was more likely for establishments of multi-unit firms than for single units.

Results on DDD are similar to those on structured management practices. Using a survey conducted on 330 large, publicly-traded firms in 2008, Brynjolfsson et al. find that output and productivity are higher for firms that depend on data to make decisions than for other firms with similar levels of investment and IT usage [58]. Using an instrumental variable method, they find that it seems likely that utilization of DDD leads to higher productivity, rather than it being the case that more productive firms are simply more able to then implement DDD.

Bloom et al. use a modified version of the WMS survey instrument’s questions on personnel management, as well as a private IT survey, accounting data, and a UK Office of National Statistics data set to show that personnel management practices are positively correlated with IT investment and productivity [59]. They find that U.S. multinationals achieve higher productivity from IT investment than do non-U.S. multinationals or non-U.S.companies broadly. The difference in IT productivity is attributed to complementary investment in personnel management practices in U.S. multinationals. Bartel et al. also find that investment in IT is accompanied by changes in personnel management practices in the valve production industry [14].

As noted above, the MOPS 2010 included two questions in “Organization” that touched upon DDD; MOPS 2015 moves these two questions to the start of the new “Data and Decision Making” component of the survey. The new module is called “Data and Decision Making” rather than “Data-driven Decision Making” so as not to lead respondents to assign value to data utilization when it is not a part of their establishment’s process.

Moving the two questions to this new section does not affect the order of the questions, but only inserts a header above these two questions, and so the comparability of the 2010 and 2015 collections should not be adversely impacted due to question order bias. The two existing questions ask if data is available to establishments and if it is being used to make decisions when available, similar to the questions asked by Brynjolfsson et al. [58].

Using the questions from the management section of the MOPS 2010, Bloom, Brynjolfsson, Foster et al. find that respondents report significant growth in data-driven monitoring practices between 2005 and 2010, which is a significant driver in overall improvement of management practices over that period, but they do not link this result to the two DDD questions [45]. Brynjolfsson and McElheran use an index constructed from the monitoring questions and the two DDD questions on the 2010 MOPS to find that larger, older plants of multi-unit firms adopt DDD earlier and to a larger extent than smaller, single-unit firms [60]. However, the single-unit firms exhibit a higher correlation with performance than similarly-sized firms that do not adopt DDD.

There are four new DDD questions on MOPS 2015. First, establishments are asked who chose what data was collected by the establishment. Second, respondents are asked to gauge how frequently four key data sources are used in the decision making process. The data sources referenced are production performance indicators from production technology or instruments, formal or informal feedback from managers, formal or informal feedback from non-managers, and outside data, which includes data from suppliers, customers, and/or outside data providers such as Federal statistical indicators. Third, MOPS 2015 also collects data on what types of decisions, namely new product design, demand forecasting, and supply chain management, are driven by data analysis and how frequently those decisions refer back to data. Fourth, respondents are asked about the reliance on predictive analytics.

As noted previously, two important components of structured management practices are targeting and monitoring. Monitoring is inherently coincidental, but targeting is a forward-looking process. The DDD section will include a fourth new question on the frequency with which decisions are made using predictive analytics such as statistical models of demand or production. This will enhance the ability of researchers to study the sophistication of the management practices being implemented by establishments. The role of DDD in predictive analysis also connects DDD and management practices with the study of uncertainty, the second new section of questions in MOPS 2015 which we turn to next.

5. Uncertainty

The final new section of the MOPS concerns uncertainty. Here we give some background that led to the eight questions in the 2015 MOPS. Like management, “uncertainty” has long been a useful explanation for economic outcomes in the popular press, policymaking, and theoretical models. Knight defined uncertainty as the inability of a person to make a forecast about an upcoming event [61]. In contrast to risk, where a person has some knowledge of an underlying probability distribution, uncertainty comes about when it is reasonably difficult to get a sense of the probability of outcomes, or even the entire outcome space. Because this definition of uncertainty involves unknown probabilities and outcomes, measuring the degree of uncertainty in the economy involves measuring the degree to which individuals are aware of unknown probability distributions.

This difficulty associated with measuring uncertainty has not stopped uncertainty from long being used as an explanation for economic outcomes. Bloom presents several key examples of the popular press suggesting that uncertainty over policy and growth has hindered investment and employment growth [62]. For example, the Federal Open Market Committee attributed a slowdown in investment to firms’ uncertainty about economic prospects in 2008, and the Chief Economist of the IMF Olivier Blanchard and then-Chair of the Council of Economic Advisors Christina Romer both cited uncertainty as a factor driving a reduction in demand in 2009. The theoretical literature allows for increasing uncertainty as an impetus for reduction in economic activity through several channels, including increasing risk premia [63] and precautionary savings [64].

Bloom examines many of the common measures of uncertainty, which include stock market volatility, GDP volatility, variation between consensus estimates and realized values of economic indicators, the Federal Reserve Bank of Philadelphia’s Survey of Professional Forecasters, and the number of appearances of the word “uncertainty” in newspaper articles or the Federal Reserve’s Beige Book [62]. A research team including Scott Baker et al. compiles indices of policy uncertainty generated from newspaper articles for the U.S., Europe, Canada, China, India, Japan, and Russia at www.policyuncertainty.com. Their index for the U.S. also includes data on expiring tax code provisions and disagreement between professional forecasters (drawn from the Survey of Professional Forecasters).

Baker et al. [65, 66] and Baker et al. [67] examine the measurement of policy uncertainty, its role in stock market fluctuations, and its potential sources, respectively. However, Jurado et al. note that the use of proxies to measure uncertainty may be useful only under a limited set of circumstances [68]. For instance, they note that “stock market volatility can change over time even if there is no change in uncertainty about economic fundamentals, if leverage changes, or if movements in risk aversion or sentiment are important drivers of asset market fluctuations”. As an alternative, they use Markov chain Monte Carlo methods to generate a measure of uncertainty from a time series consisting of 132 mostly macroeconomic variables and 147 financial variables.

The aforementioned proxies of policy uncertainty have been widely used in finance, and have been presented in congressional and Federal Reserve testimony. Bloom uses stock market volatility to show that bad news uncertainty shocks are associated with reductions in hiring and investment [69]. Similarly et al. use deviations in stock returns to show that uncertainty reduces investment [70]. If one takes the view, as in Bloom et al. [49], that management is a technology, then adoption of management practices can be viewed as a form of investment. However, the relationship between uncertainty and adoption of structured management practices has been largely untested to this point.

Several surveys by central banks take the approach of directly asking households and businesses for their expectations over various economic outcomes. The Bank of Japan’s TANKAN is sent out to 210,000 large firms quarterly and can be answered by mail or online.Firms are asked to judge their views of business conditions, inventories, capacity, employment, finances, and other topics at the present, and then asked to give annual projections on sales, exports, exchange rates, profits, income, investment, and inflation. Similarly, The Bank of Italy’s Survey on Inflation and Growth Expectations is issued annually and manufacturing firms are asked about investment levels for the current year, which includes a partial forecast. D’Aurizio and Iezzi use these qualitative responses to build a forecasting model of investment [71]. The Federal Reserve Bank of Philadelphia also runs a monthly Business Outlook Survey (BOS) in which 100–125 manufacturing firms are asked only if certain economic indicators (orders, shipments, prices, employees, etc.) are expected to increase, decrease, or remain unchanged within the next six months.

The Ifo Institute Center for Economic Studies in Munich has run the Ifo Business Climate Survey (IFO-BCS) that surveys between 2,500 and 5,000 German products (which cover 2,000–4,000 continuing firms) on a monthly basis with consistent data running back to 1980. Respondents are asked to characterize their expectations of business conditions as “more favorable”, “unchanged”, or “more unfavorable”.Bachmann et al. use both the BOS and IFO-BCS to show that adverse supply shocks tend to increase uncertainty, but uncertainty in the absence of shocks have only limited adverse effects on real activity [72].

The Federal Reserve Bank of Atlanta, in partnership with Steven Davis of the University of Chicago Booth School of Business and Nicholas Bloom of Stanford University, has created the Decision Maker Survey to measure firms’ year-ahead expectations and associated uncertainties regarding changes in their costs, prices, profit margins, level of employment, capital investment, and sales revenue. The survey panel consists of a national sample of firms representing every sector of the economy (with the exception of agriculture and government) and a broad range of firm sizes. In addition, the Federal Reserve Bank of Atlanta runs the Business Inflation Expectations survey, which asks 300 firms monthly to assign probabilities to six potential outcomes for inflation over the next twelve months, and asks a pair of questions on its biannual Small Business Survey (SBS) on uncertainty. The SBS covers firms with fewer than 500 employees and asks respondents whether “uncertainty” as a broad concept is having a larger or smaller impact on the firm’s decision making relative to six months prior, and then asks respondents to cite the primary source of uncertainty.

The Census Bureau’s annual Business R&D and Innovation Survey (BRDIS) asks respondents for firm-level forecasts of R&D expenditure for the year following the coverage year (which is the year in which the survey is completed). The BRDIS also includes similarly structured questions on forecasted foreign and domestic R&D expenditure and the amount of R&D expenditure paid for by others.

The link between management and uncertainty is discussed in some of the theoretical literature, including Adam which uses management as a motivator for limited capacity for information processing [7]. If managers are responsible for gathering and processing information and setting targets, then managers are responsible, in some sense, for monitoring uncertainty. Do better management practices and more data-driven decision making lead to better forecasts and reduced uncertainty? Does the presence of uncertainty increase investment in management because of this effect? Or does uncertainty reduce investment in management practices due to precautionary savings on the part of the establishment? Limited research exists to this point on the role of management in the quality of forecasts, but Ben-David et al. find that executives are often incorrect with regards to their forecasts of stock market distributions [73].

MOPS 2015 includes eight new questions on uncertainty. There are two uncertainty questions on each of the following four subjects: shipments, capital expenditure, employment, and the cost of materials, parts, containers, and packaging. The first question for each subject asks for an estimate of the value of the variable in question in 2015 as well as a partial forecast of 2016, which will be roughly one-third complete at the time that respondents receive the survey. The latter portion of these questions is in the vein of the Italian Survey on Inflation and Growth Expectations, while the former allows for a measure of the measurement error of the respondents relative to the ASM. Note that the questions on employment ask for employment as of March 12 for consistency with the ASM. Since MOPS 2015 was mailed on April 28, 2016, the question on employment in 2015 and 2016 will not include a forecasting component.

The second question asks respondents for five points of their possible distribution of possible outcomes at the plant for 2017 (lowest, low, middle, high, and highest) and the likelihood that they would assign to each outcome. Taken together these questions can be used to estimate the moments of the distributions of the variables in question, which provides a much richer measure of uncertainty than the proxies outlined above. Because this set of questions is somewhat abstract, the section is preceded by instructions with an example of how a hypothetical respondent might fill out a pair of uncertainty questions.

6. Conclusions

Management has long been used as a residual in the explanation of why performance differs across firms and establishments. While business schools and the popular press have emphasized the importance of particular management practices, only in the last ten years have economists devoted significant empirical study to management practices. As the largest single study of management practices and the first large-scale study of management in the United States, the MOPS is at the center of this burgeoning field of research.

The research team (external researchers and Census researchers) published the first detailed results of the MOPS 2010 data in a CES working paper. Bloom et al. report findings that are consistent with the earlier work from the WMS [45]. Firms that adopt more of the structured management practices related to monitoring, targeting, and incentives are more productive, more profitable, and grow faster than firms with lower levels of structured management practice adoption. They also find that there are high levels of dispersion in structured management practice adoption, with higher levels of adoption being found in the South and Midwest, in larger establishments, in establishments of large firms, in exporting establishments, and in establishments with more educated employees. Finally, the authors find that establishments generally report higher levels of implementation of structured management practices since 2005. They also find that firms with higher management scores generally have higher rates of innovation, invest more heavily in IT, and have higher stock market valuations [74].

The second collection of the MOPS will enable us to better understand the dynamics of management practices. Moreover, the expanded version of the MOPS includes questions on two new subjects related to management: data and decision making and uncertainty. Because management is concerned at least in part with monitoring and setting forecasts, data collection and usage is an important complement to structured management practices. Furthermore, since targeting is at least in part forward-looking, structured management practices must also be related to the study of uncertainty. With its sixteen new questions (four on background, four on DDD, and eight on uncertainty), it will be exciting to see how the MOPS 2015 adds to our understanding of management practices in the U.S.

Supplementary data

The supplementary files are available to download from https://dx-doi-org.web.bisu.edu.cn/10.3233/JEM-170441.