We present LFCMV, a Bayesian file linking methodology designed to link records using continuous matching variables in situations where we do not expect values of these matching variables to agree exactly across matched pairs. The method involves a linking model for the distance between the matching variables of records in one file and the matching variables of their linked records in the second. This linking model is conditional on a vector indicating the links. We specify a mixture model for the distance component of the linking model, as this latent structure allows the distance between matching variables in linked pairs to vary across types of linked pairs. Finally, we specify a model for the linking vector. We describe the Gibbs sampling algorithm for sampling from the posterior distribution of this linkage model and use artificial data to illustrate model performance. We also introduce a linking application using public survey information and data from the U.S. Census of Manufactures and use
LFCMV to link the records.
-
USING IMPUTATION TECHNIQUES TO EVALUATE STOPPING RULES IN ADAPTIVE SURVEY DESIGN
October 2014
Working Paper Number:
CES-14-40
Adaptive Design methods for social surveys utilize the information from the data as it is collected to make decisions about the sampling design. In some cases, the decision is either to continue or stop the data collection. We evaluate this decision by proposing measures to compare the collected data with follow-up samples. The options are assessed by imputation of the nonrespondents under different missingness scenarios, including Missing Not at Random. The variation in the utility measures is compared to the cost induced by the follow-up sample sizes. We apply the proposed method to the 2007 U.S. Census of Manufacturers.
View Full
Paper PDF
-
Squeezing More Out of Your Data: Business Record Linkage with Python
November 2018
Working Paper Number:
CES-18-46
Integrating data from different sources has become a fundamental component of modern data analytics. Record linkage methods represent an important class of tools for accomplishing such integration. In the absence of common disambiguated identifiers, researchers often must resort to ''fuzzy" matching, which allows imprecision in the characteristics used to identify common entities across dfferent datasets. While the record linkage literature has identified numerous individually useful fuzzy matching techniques, there is little consensus on a way to integrate those techniques within a
single framework. To this end, we introduce the Multiple Algorithm Matching for Better Analytics (MAMBA), an easy-to-use, flexible, scalable, and transparent software platform for business record linkage applications using Census microdata. MAMBA leverages multiple string comparators to assess the similarity of records using a machine learning algorithm to disambiguate matches. This software represents a transparent tool for researchers seeking to link external business data to the Census Business Register files.
View Full
Paper PDF
-
Plant-Level Productivity and Imputation of Missing Data in the Census of Manufactures
January 2011
Working Paper Number:
CES-11-02
In the U.S. Census of Manufactures, the Census Bureau imputes missing values using a combination of mean imputation, ratio imputation, and conditional mean imputation. It is wellknown that imputations based on these methods can result in underestimation of variability and potential bias in multivariate inferences. We show that this appears to be the case for the existing imputations in the Census of Manufactures. We then present an alternative strategy for handling the missing data based on multiple imputation. Specifically, we impute missing values via sequences of classification and regression trees, which offer a computationally straightforward and flexible approach for semi-automatic, large-scale multiple imputation. We also present an approach to evaluating these imputations based on posterior predictive checks. We use the multiple imputations, and the imputations currently employed by the Census Bureau, to estimate production function parameters and productivity dispersions. The results suggest that the two approaches provide quite different answers about productivity.
View Full
Paper PDF
-
Exploring New Ways to Classify Industries for Energy Analysis and Modeling
November 2022
Working Paper Number:
CES-22-49
Combustion, other emitting processes and fossil energy use outside the power sector have become urgent concerns given the United States' commitment to achieving net-zero greenhouse gas emissions by 2050. Industry is an important end user of energy and relies on fossil fuels used directly for process heating and as feedstocks for a diverse range of applications. Fuel and energy use by industry is heterogeneous, meaning even a single product group can vary broadly in its production routes and associated energy use. In the United States, the North American Industry Classification System (NAICS) serves as the standard for statistical data collection and reporting. In turn, data based on NAICS are the foundation of most United States energy modeling. Thus, the effectiveness of NAICS at representing energy use is a limiting condition for current
expansive planning to improve energy efficiency and alternatives to fossil fuels in industry. Facility-level data could be used to build more detail into heterogeneous sectors and thus supplement data from Bureau of the Census and U.S Energy Information Administration reporting at NAICS code levels but are scarce. This work explores alternative classification schemes for industry based on energy use characteristics and validates an approach to estimate facility-level energy use from publicly available greenhouse gas emissions data from the U.S. Environmental Protection Agency (EPA). The approaches in this study can facilitate understanding of current, as well as possible future, energy demand.
First, current approaches to the construction of industrial taxonomies are summarized along with their usefulness for industrial energy modeling. Unsupervised machine learning techniques are then used to detect clusters in data reported from the U.S. Department of Energy's Industrial Assessment Center program. Clusters of Industrial Assessment Center data show similar levels of correlation between energy use and explanatory variables as three-digit NAICS codes. Interestingly, the clusters each include a large cross section of NAICS codes, which lends additional support to the idea that NAICS may not be particularly suited for correlation between energy use and the variables studied. Fewer clusters are needed for the same level of correlation as shown in NAICS codes. Initial assessment shows a reasonable level of separation using support vector machines with higher than 80% accuracy, so machine learning approaches may be promising for further analysis. The IAC data is focused on smaller and medium-sized facilities and is biased toward higher energy users for a given facility type. Cladistics, an approach for classification developed in biology, is adapted to energy and process characteristics of industries. Cladistics applied to industrial systems seeks to understand the progression of organizations and technology as a type of evolution, wherein traits are inherited from previous systems but evolve due to the emergence of inventions and variations and a selection process driven by adaptation to pressures and favorable outcomes. A cladogram is presented for evolutionary directions in the iron and steel sector. Cladograms are a promising tool for constructing scenarios and summarizing directions of sectoral innovation.
The cladogram of iron and steel is based on the drivers of energy use in the sector. Phylogenetic inference is similar to machine learning approaches as it is based on a machine-led search of the solution space, therefore avoiding some of the subjectivity of other classification systems. Our prototype approach for constructing an industry cladogram is based on process characteristics according to the innovation framework derived from Schumpeter to capture evolution in a given sector. The resulting cladogram represents a snapshot in time based on detailed study of process characteristics. This work could be an important tool for the design of scenarios for more detailed modeling. Cladograms reveal groupings of emerging or dominant processes and their implications in a way that may be helpful for policymakers and entrepreneurs, allowing them to see the larger picture, other good ideas, or competitors. Constructing a cladogram could be a good first step to analysis of many industries (e.g. nitrogenous fertilizer production, ethyl alcohol manufacturing), to understand their heterogeneity, emerging trends, and coherent groupings of related innovations.
Finally, validation is performed for facility-level energy estimates from the EPA Greenhouse Gas Reporting Program. Facility-level data availability continues to be a major challenge for industrial modeling. The method outlined by (McMillan et al. 2016; McMillan and Ruth 2019) allows estimating of facility level energy use based on mandatory greenhouse gas reporting. The validation provided here is an important step for further use of this data for industrial energy modeling.
View Full
Paper PDF
-
Distribution Preserving Statistical Disclosure Limitation
September 2006
Working Paper Number:
tp-2006-04
One approach to limiting disclosure risk in public-use microdata is to release multiply-imputed,
partially synthetic data sets. These are data on actual respondents, but with confidential data
replaced by multiply-imputed synthetic values. A mis-specified imputation model can invalidate
inferences because the distribution of synthetic data is completely determined by the model used
to generate them. We present two practical methods of generating synthetic values when the imputer
has only limited information about the true data generating process. One is applicable when
the true likelihood is known up to a monotone transformation. The second requires only limited
knowledge of the true likelihood, but nevertheless preserves the conditional distribution of the confidential
data, up to sampling error, on arbitrary subdomains. Our method maximizes data utility
and minimizes incremental disclosure risk up to posterior uncertainty in the imputation model and
sampling error in the estimated transformation. We validate the approach with a simulation and
application to a large linked employer-employee database.
View Full
Paper PDF
-
Simultaneous Edit-Imputation for Continuous Microdata
December 2015
Working Paper Number:
CES-15-44
Many statistical organizations collect data that are expected to satisfy linear constraints; as examples, component variables should sum to total variables, and ratios of pairs of variables should be bounded by expert-specified constants. When reported data violate constraints, organizations identify and replace values potentially in error in a process known as edit-imputation. To date, most approaches separate the error localization and imputation steps, typically using optimization methods to identify the variables to change followed by hot deck imputation. We present an approach that fully integrates editing and imputation for continuous microdata under linear constraints. Our approach relies on a Bayesian hierarchical model that includes (i) a flexible joint probability model for the underlying true values of the data with support only on the set of values that satisfy all editing constraints, (ii) a model for latent indicators of the variables that are in error, and (iii) a model for the reported responses for variables in error. We illustrate the potential advantages of the Bayesian editing approach over existing approaches using simulation studies. We apply the model to edit faulty data from the 2007 U.S. Census of Manufactures. Supplementary materials for this article are available online.
View Full
Paper PDF
-
Empirical Distribution of the Plant-Level Components of Energy and Carbon Intensity at the Six-digit NAICS Level Using a Modified KAYA Identity
September 2024
Working Paper Number:
CES-24-46
Three basic pillars of industry-level decarbonization are energy efficiency, decarbonization of energy sources, and electrification. This paper provides estimates of a decomposition of these three components of carbon emissions by industry: energy intensity, carbon intensity of energy, and energy (fuel) mix. These estimates are constructed at the six-digit NAICS level from non-public, plant-level data collected by the Census Bureau. Four quintiles of the distribution of each of the three components are constructed, using multiple imputation (MI) to deal with non-reported energy variables in the Census data. MI allows the estimates to avoid non-reporting bias. MI also allows more six-digit NAICS to be estimated under Census non-disclosure rules, since dropping non-reported observations may have reduced the sample sizes unnecessarily. The estimates show wide variation in each of these three components of emissions (intensity) and provide a first empirical look into the plant-level variation that underlies carbon emissions.
View Full
Paper PDF
-
IMPROVING THE SYNTHETIC LONGITUDINAL BUSINESS DATABASE
February 2014
Working Paper Number:
CES-14-12
In most countries, national statistical agencies do not release establishment-level business microdata, because doing so represents too large a risk to establishments' confidentiality. Agencies potentially can manage these risks by releasing synthetic microdata, i.e., individual establishment records simulated from statistical models de- signed to mimic the joint distribution of the underlying observed data. Previously, we used this approach to generate a public-use version'now available for public use'of the U. S. Census Bureau's Longitudinal Business Database (LBD), a longitudinal cen- sus of establishments dating back to 1976. While the synthetic LBD has proven to be a useful product, we now seek to improve and expand it by using new synthesis models and adding features. This article describes our efforts to create the second generation of the SynLBD, including synthesis procedures that we believe could be replicated in other contexts.
View Full
Paper PDF
-
A Search and Learning Model of Export Dynamics
August 2021
Working Paper Number:
CES-21-17
Exporting abroad is much harder than selling at home, and overcoming hurdles to exporting takes time. Our goal is to identify specific barriers to exporting and to measure their importance. We develop a model of firm-level export dynamics that features costly customer search, network effects in finding buyers, and learning about product appeal. Fitting the model to customs records of U.S. imports of manufactures from Colombia we replicate patterns of exporter maturation. A potentially valuable intangible asset of a firm is its customer base and knowledge of a market. Our model delivers some striking estimates of what such assets are worth. Averaging across active exporters, the loss from total market amnesia (losing its current U.S. customer base along with its accumulated knowledge of product appeal) is US$ 3.4 million, about 34 percent of the value of exporting overall. About half is the loss of future sales to existing customers while the rest is the cost of relearning its appeal in the market and reestablishing visibility as an exporter. Given the importance of search, learning, and visibility, the 5-year response of total export sales to an exchange rate shock exceeds the 1-year response by about 40 percent.
View Full
Paper PDF
-
Recall and Response: Relationship Adjustments to Adverse Information Shocks
March 2020
Working Paper Number:
CES-20-13R
How resilient are U.S. buyer-foreign supplier relationships to new information about product defects? We construct a novel dataset of U.S. consumer-product recalls sourced from foreign suppliers between 1995 and 2013. Using an event-study approach, we find that compared to control relationships, buyers that experience recalls temporarily reduce their probability of trading with the suppliers of the recalled products by 17%. The reduction is much larger for new than established buyer'supplier relationships. Buyers that experience a recall are more likely to add other suppliers to their portfolios, diversifying supplier-specific risk in the aftermath of a recall; this effect, too, is larger for buyers impacted by recalls in new relationships. There is a long lag ' up to two years ' before diversification, consistent with a high cost of establishing new relationships.
View Full
Paper PDF