-
Gradient Boosting to Address Statistical Problems Arising from Non-Linkage of Census Bureau Datasets
June 2024
Working Paper Number:
CES-24-27
This article introduces the twangRDC package, which contains functions to address non-linkage in US Census Bureau datasets. The Census Bureau's Person Identification Validation System facilitates data linkage by assigning unique person identifiers to federal, third party, decennial census, and survey data. Not all records in these datasets can be linked to the reference file and as such not all records will be assigned an identifier. This article is a tutorial for using the twangRDC to generate nonresponse weights to account for non-linkage of person records across US Census Bureau datasets.
View Full
Paper PDF
-
Improving Patent Assignee-Firm Bridge with Web Search Results
August 2022
Working Paper Number:
CES-22-31
This paper constructs a patent assignee-firm longitudinal bridge between U.S. patent assignees and firms using firm-level administrative data from the U.S. Census Bureau. We match granted patents applied between 1976 and 2016 to the U.S. firms recorded in the Longitudinal Business Database (LBD) in the Census Bureau. Building on existing algorithms in the literature, we first use the assignee name, address (state and city), and year information to link the two datasets. We then introduce a novel search-aided algorithm that significantly improves the matching results by 7% and 2.9% at the patent and the assignee level, respectively. Overall, we are able to match 88.2% and 80.1% of all U.S. patents and assignees respectively. We contribute to the existing literature by 1) improving the match rates and quality with the web search-aided algorithm, and 2) providing the longest and longitudinally consistent crosswalk between patent assignees and LBD firms.
View Full
Paper PDF
-
Identifying U.S. Merchandise Traders: Integrating Customs Transactions with Business Administrative Data
September 2020
Working Paper Number:
CES-20-28
This paper describes the construction of the Longitudinal Firm Trade Transactions Database (LFTTD) enabling the identification of merchandise traders - exporters and importers - in the U.S. Census Bureau's Business Register (BR). The LFTTD links merchandise export and import transactions from customs declaration forms to the BR beginning in 1992 through the present. We employ a combination of deterministic and probabilistic matching algorithms to assign a unique firm identifier in the BR to a merchandise export or import transaction record. On average, we match 89 percent of export and import values to a firm identifier. In 1992, we match 79 (88) percent of export (import) value; in 2017, we match 92 (96) percent of export (import) value. Trade transactions in year t are matched to years between 1976 and t+1 of the BR. On average, 94 percent of the trade value matches to a firm in year t of the BR. The LFTTD provides the most comprehensive identification of and the foundation for the analysis of goods trading firms in the U.S. economy.
View Full
Paper PDF
-
Matching State Business Registration Records
to Census Business Data
January 2020
Working Paper Number:
CES-20-03
We describe our methodology and results from matching state Business Registration Records (BRR) to Census business data. We use data from Massachusetts and California to develop methods and preliminary results that could be used to guide matching data for additional states. We obtain matches to Census business records for 45% of the Massachusetts BRR records and 40% of the California BRR records. We find higher match rates for incorporated businesses and businesses with higher startup-quality scores as assigned in Guzman and Stern (2018). Clerical reviews show that using relatively strict matching on address is important for match accuracy, while results are less sensitive to name matching strictness. Among matched BRR records, the modal timing of the first match to the BR is in the year in which the BRR record was filed. We use two sets of software to identify matches: SAS DQ Match and a machine-learning algorithm described in Cuffe and Goldschlag (2018). We find preliminary evidence that while the ML-based method yields more match results, SAS DQ tends to result in higher accuracy rates. To conclude, we provide suggestions on how to proceed with matching other states' data in light of our findings using these two states.
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
-
Assessing Coverage and Quality of the 2007 Prototype Census Kidlink Database
September 2015
Working Paper Number:
carra-2015-07
The Census Bureau is conducting research to expand the use of administrative records data in censuses and surveys to decrease respondent burden and reduce costs while improving data quality. Much of this research (e.g., Rastogi and O''Hara (2012), Luque and Bhaskar (2014)) hinges on the ability to integrate multiple data sources by linking individuals across files. One of the Census Bureau's record linkage methodologies for data integration is the Person Identification Validation System or PVS. PVS assigns anonymous and unique IDs (Protected Identification Keys or PIKs) that serve as linkage keys across files. Prior research showed that integrating 'known associates' information into PVS's reference files could potentially enhance PVS's PIK assignment rates. The term 'known associates' refers to people that are likely to be associated with each other because of a known common link (such as family relationships or people sharing a common address), and thus, to be observed together in different files. One of the results from this prior research was the creation of the 2007 Census Kidlink file, a child-level file linking a child's Social Security Number (SSN) record to the SSN of those identified as the child's parents. In this paper, we examine to what extent the 2007 Census Kidlink methodology was able to link parents SSNs to children SSN records, and also evaluate the quality of those links. We find that in approximately 80 percent of cases, at least one parent was linked to the child's record. Younger children and noncitizens have a higher percentage of cases where neither parent could be linked to the child. Using 2007 tax data as a benchmark, our quality evaluation results indicate that in at least 90 percent of the cases, the parent-child link agreed with those found in the tax data. Based on our findings, we propose improvements to the 2007 Kidlink methodology to increase child-parent links, and discuss how the creation of the file could be operationalized moving forward.
View Full
Paper PDF
-
Person Matching in Historical Files using the Census Bureau's Person Validation System
September 2014
Working Paper Number:
carra-2014-11
The recent release of the 1940 Census manuscripts enables the creation of longitudinal data spanning the whole of the twentieth century. Linked historical and contemporary data would allow unprecedented analyses of the causes and consequences of health, demographic, and economic change. The Census Bureau is uniquely equipped to provide high quality linkages of person records across datasets. This paper summarizes the linkage techniques employed by the Census Bureau and discusses utilization of these techniques to append protected identification keys to the 1940 Census.
View Full
Paper PDF
-
Estimating Record Linkage False Match Rate for the Person Identification Validation System
July 2014
Working Paper Number:
carra-2014-02
The Census Bureau Person Identification Validation System (PVS) assigns unique person identifiers to federal, commercial, census, and survey data to facilitate linkages across files. PVS uses probabilistic matching to assign a unique Census Bureau identifier for each person. This paper presents a method to measure the false match rate in PVS following the approach of Belin and Rubin (1995). The Belin and Rubin methodology requires truth data to estimate a mixture model. The parameters from the mixture model are used to obtain point estimates of the false match rate for each of the PVS search modules. The truth data requirement is satisfied by the unique access the Census Bureau has to high quality name, date of birth, address and Social Security (SSN) data. Truth data are quickly created for the Belin and Rubin model and do not involve a clerical review process. These truth data are used to create estimates for the Belin and Rubin parameters, making the approach more feasible. Both observed and modeled false match rates are computed for all search modules in federal administrative records data and commercial data.
View Full
Paper PDF
-
The Person Identification Validation System (PVS): Applying the Center for Administrative Records Research and Applications' (CARRA) Record Linkage Software
July 2014
Working Paper Number:
carra-2014-01
The Census Bureau's Person Identification Validation System (PVS) assigns unique person identifiers to federal, commercial, census, and survey data to facilitate linkages across and within files. PVS uses probabilistic matching to assign a unique Census Bureau identifier for each person. The PVS matches incoming files to reference files created with data from the Social Security Administration (SSA) Numerical Identification file, and SSA data with addresses obtained from federal files. This paper describes the PVS methodology from editing input data to creating the final file.
View Full
Paper PDF
-
Comparison of Survey, Federal, and Commercial Address Data Quality
June 2014
Working Paper Number:
carra-2014-06
This report summarizes matching of survey, commercial, and administrative records housing units to the Census Bureau Master Address File (MAF). We document overall MAF match rates in each data set and evaluate differences in match rates across a variety of housing characteristics. Results show that over 90 percent of records in survey data from the American Housing Survey (AHS) match to the MAF. Commercial data from CoreLogic matches at much lower rates, in part due to missing address information and poor match rates for multi-unit buildings. MAF match rates for administrative records from the Department of Housing and Urban Development are also high, and open the possibility of using this information in surveys such as the AHS.
View Full
Paper PDF
