Click here to download a PDF version of this documentlast updated on 2025-11-13
Click here to download a PDF version of this document
This document provides an overview of the sources and processing applied to each data series within the Opportunity Insights Economic Tracker. The documentation is organized sequentially by series in the tracker, then broken down into categories of information describing each series, its source data, and our processing steps.
You can refer to additional documentation published by Opportunity Insights for complementary information:
Please note that both the data and this data documentation are updated regularly and that the following information is subject to change.
Summary: Aggregated and anonymized purchase data from consumer credit and debit card spending. Spending is reported based on the ZIP code where the cardholder lives, not the ZIP code where transactions occurred.
Data Source: Affinity Solutions
Update Frequency: Weekly
Date Range: January 13, 2020 until the most recent date available.
Data Frequency: Data is daily until June 5, 2022, presented as a 7-day lookback moving average. Since June 5, 2022 we only receive weekly data on consumer spending, and the data is presented as weekly data points.
Indexing Period: January 6 to February 2, 2020
Indexing Type: Seasonally adjusted change since January 2020. We calculate the change relative to the January index period: 2019 data is indexed relative to January 2019 (January 7, 2019 to February 3, 2019), data in 2020 onward is indexed relative to January 2020 (January 6, 2020 to February 2, 2020). We then seasonally adjust by dividing by the indexed 2019 value, which represents the difference between the change since January 2020 compared to the change since January observed since 2019. We account for differences in the dates of federal holidays between 2019 and 2020 by shifting the 2019 reference data to align the holidays before performing the year-over-year division.
Geographies: National, State, County, Metro
Breakdowns:
By Industry. Industries are constructed by grouping merchant codes that are used by Affinity Solutions to identify the category of merchant and merchant activity.
By Consumer Zip Code Income. Transactions are linked to zip codes where the consumer lives and zip codes are classified into income categories based on measurements of median household income and population provided by the American Community Survey (2014 - 2018).
Data masking: For the state-level breakdowns by income quartile and the county-level data, we mask locations with average daily spending < $70,000 in January 2019. The raw data contains discontinuous breaks caused by entry or exit of credit card providers from the sample: counties with multiple structural breaks are dropped from the sample. Additionally, Affinity Solutions suppresses any cut of the data with fewer than five transactions. For more details refer to the accompanying paper.
Notes: We require at least 3 weeks of data in order to reliably identify and correct discontinuous breaks caused by entry or exit of credit card providers from the sample. The most recent 3 weeks of data are therefore marked ‘provisional’ and are subject to non-negligible changes as new data is posted. For breaks found prior to the last 3 weeks, we correct for it using a method outlined in the paper. Otherwise we substitute the national mean for more recent breaks while we gather enough data to implement the corrections outlined in the paper. We typically only allow series to have one significant break, however, in the case of areas with large populations we relax this rule.
Supplemental files:
share_jan2019: the share of weekly total spending from
January 7 to February 3, 2019 by income quartile.share_jan2020: the share of weekly total spending from
January 6 to February 2, 2020 by income quartile.share_jan2020: the share of weekly total spending from
January 6 to February 2, 2020 by MCC group.share_decline_covidfirstwave: the share of the decline
in weekly total spending in the first wave of the pandemic by MCC group.
To measure the decline, we first take the sum of total spending in four
periods: from January 7 to February 3, 2019, from March 25 to April 14,
2019, from January 6 to February 2, 2020, and from March 30 to April 20,
2020. Then, the decline is calculated as the change in spending between
April 2019 and April 2020, minus the change in spending between January
2019 and January 2020. This measures the decline in total spending
during the first wave of the COVID-19 pandemic, accounting for the
normal decline in spending across the same months in 2019.daily_spend_19_all and daily_spend_19_q#:
daily total spending indexed to total spending from January 7 to
February 3, 2019 by income quartile, not seasonally adjusted.Summary: Small business transactions and revenue data aggregated from several credit card processors. Transactions and revenue are reported based on the county or ZIP code where the business is located.
Data Source: Womply
Update Frequency: Weekly
Date Range: January 15, 2020 until the most recent date available.
Data Frequency: Weekly
Indexing Period: January 4 to January 31, 2020
Indexing Type: Seasonally adjusted change since January 2020. We calculate the change relative to the January index period: 2019 data is indexed relative to January 2019, data in 2020 onward is indexed relative to January 2020. We then seasonally adjust by dividing by the indexed 2019 value, which represents the difference between the change since January 2020 compared to the change since January observed since 2019.
In all series we adjust for holidays while seasonally adjusting a given series to ensure that we are comparing weeks containing a significant holiday to a corresponding week containing that holiday. This adjustment has three components: (1) weeks with the same holiday are always compared to each other; (2) weeks before the same holiday are always compared to each other; and (3) weeks after holidays are always compared to each other. Thus, even if a holiday does not fall in the same week in the year being normed and the comparison year, weeks will be aligned when compared to each other.
Geographies: National, State, County, Metro
Breakdowns:
Industry, by 2-digit NAICS.
Business Zip Code Income. Transactions are linked to ZIP codes where the business is located and ZIP codes are classified into income categories based on measurements of median household income and population provided by the American Community Survey (2014 - 2018).
Data Masking:
The data we receive from Womply is restricted to businesses that have an annual revenue that is less than the SBA thresholds by industry, and have an average revenue that is within 3 standard deviations of the state average.
We omit counties that don’t have a minimum of 3 businesses operating in the first week of January 2020, January 2021, and January 2022. For the county-level series, we mask any counties for which revenue is less than $250,000 during the indexing period (January 4 to 31, 2020): such counties are incorporated into state-level or national-level aggregates but are not reported isolated in the county-level data.
To reduce outliers, we manually exclude some state x industry breakdowns that present extreme variation from our state and national level calculations, as well as a small number of counties that demonstrate extreme variation.
Notes: Subnational breakdowns by High/Middle/Low income ZIP codes have been temporarily removed since the August 21st 2020 update due to revisions in the structure of the raw data we receive. We hope to add them back to the OI Economic Tracker soon.
Supplemental files:
revenue_all_apr2020: Percent change in net revenue for
small businesses from January 4 to 31, 2020 to March 23 to April 12,
2020, seasonally adjusted.revenue_all_jul2020: Percent change in net revenue for
small businesses from January 4 to 31, 2020 to June 29 to July 26, 2020,
seasonally adjusted.Summary: Number of small businesses open, as defined by having had at least one transaction in the previous 3 days.
Data Source: Womply
Update Frequency: Weekly
Date Range: January 15, 2020 until the most recent date available.
Data Frequency: Weekly
Indexing Period: January 4 to 31, 2020
Indexing Type: Seasonally adjusted change since January 2020. We calculate the change relative to the January index period: 2019 data is indexed relative to January 2019, data in 2020 onward is indexed relative to January 2020. We then seasonally adjust by dividing by the indexed 2019 value, which represents the difference between the change since January 2020 compared to the change since January observed since 2019.
In all series we adjust for holidays while seasonally adjusting a given series to ensure that we are comparing weeks containing a significant holiday to a corresponding week containing that holiday. This adjustment has three components: (1) weeks with the same holiday are always compared to each other; (2) weeks before the same holiday are always compared to each other; and (3) weeks after holidays are always compared to each other. Thus, even if a holiday does not fall in the same week in the year being normed and the comparison year, weeks will be aligned when compared to each other.
Geographies: National, State, County, Metro
Breakdowns:
Industry, by 2-digit NAICS.
Business Zip Code Income. Transactions are linked to ZIP codes where the business is located and ZIP codes are classified into income categories based on measurements of median household income and population provided by the American Community Survey (2014 - 2018).
Data Masking:
The data we receive from Womply is restricted to businesses that have an annual revenue that is less than the SBA thresholds by industry, and have an average revenue that is within 3 standard deviations of the state average.
We omit counties that don’t have a minimum of 3 businesses operating in the first week of January 2020, January 2021 and January 2022. For the county-level series, we mask any counties for which revenue is less than $250,000 during the indexing period (January 4 to 31 2020): such counties are incorporated into state-level or national-level aggregates but are not reported isolated in the county-level data.
To reduce outliers, we manually exclude some state x industry breakdowns that present extreme variation from our state and national level calculations, as well as a small number of counties that demonstrate extreme variation.
Notes: Subnational breakdowns by High/Middle/Low income ZIP codes have been temporarily removed since the August 21st 2020 update due to revisions in the structure of the raw data we receive. We hope to add them back to the OI Economic Tracker soon.
Summary: Weekly count of new job postings, sourced from over 40,000 online job boards. New job postings are defined as those that have not had a duplicate posting for at least 60 days prior.
Data Source: Lightcast (formerly known as Burning Glass Technologies)
Update Frequency: Weekly
Date Range: January 17, 2020 until the most recent date available.
Data Frequency: Weekly data points, with each week ending on Friday.
Indexing Period: January 4 to 31, 2020
Indexing Type: Change relative to the January 2020 index period, not seasonally adjusted.
Geographies: National, State, County, Metro.
Breakdowns:
Industry, by NAICS supersector.
Education Requirement, by ONET Jobzone’s Education Requirement Classification.
Data Masking: In order to avoid extreme outliers, we calculate a cutoff of one standard deviation above the 97th percentile of the state-level data for each variable and mask values that exceed this threshold. Additionally, at the county level, only subgroup data for the 200 largest counties is able to be disclosed for firm data privacy reasons. For the remaining counties’ subgroups, all values are imputed from the share of state postings that are made up of a given subgrouping multiplied by the number of county postings in total.
Supplemental files:
share_jan2020: the share of job postings by industry
(2-digit NAICS) in the period from January 4 to 31, 2020.Summary: Please note we are planning the release of a new, updated, and overhauled employment series in the coming weeks. This series will much better handle entry and exit from the underlying sample as well as improve on other biases in the current series. As such we advise that users wait for the release of the new and updated data in order before using the Employment series so that users have the most accurate information available. Number of active employees, aggregating information from multiple data providers. This series is based on firm-level payroll data from Paychex and Intuit.
Update Frequency: Weekly
Date Range: January 15, 2020 until the most recent date available. The most recent date available for the full series depends on the combination of Paychex and Intuit data.
Data Frequency: Weekly
Indexing Period: January 4 to 31, 2020
Indexing Type: Change relative to the January 2020 index period, not seasonally adjusted.
Geographies: National, State, County, Metro
Breakdowns:
Wage.
Industry, by NAICS supersector.
Industry, by NAICS sector.
Data Masking: As the employment series is a composite series, each of its component series have their own masking standards that in aggregate determine masking for the series.
In the Paychex series, we reduce the weight of cells in which we detect firm entry/exit over time. In each county x industry (two-digit NAICS code) x firm size x wage quartile cell, we compute the change in employment relative to January 4 to 31, 2020, and the change in employment relative to July 1 to 31, 2020. For county x industry x firm size x wage quartile cells between January 2020 and the end of the series, we reduce the weight we place on the series if we observe changes in employment that indicate firm entry or exit.
The difference in weighting between small and large cells is to account for large amounts of small firm births, particularly in the second half of 2020, which played a strong role in the economic recovery from the pandemic.
In the Intuit series, we do not make any sample restrictions.
Supplemental files:
emp_incq1_apr2020: Change in employment level for all
workers in the Earnin sample (low-wage) from January 2020 to April
2020.emp_incq1_jul2020: Change in employment level for all
workers in the Earnin sample (low-wage), from January 2020 to July
2020.Summary: Weekly unemployment insurance claims counts and rates (as a share of the 2019 labor force) for all states, as well as initial unemployment insurance claims for select counties where the data is publicly available.
Data Source: State-level and national statistics are reported by the U.S. Department of Labor.
The county-level series is only available for states whose respective state agencies publish county level data:
Update Frequency: Weekly (where available, in the case of county-level data)
Date Range: January 18, 2020 until the most recent date available.
Data Frequency: Weekly data points, with each week ending on Saturday.
Note that county-level claims in California, Georgia, Kentucky, and Illinois are reported at the monthly level and imputed to weekly data points for the county-level series. For more information about the imputation methodology, see the accompanying paper
Indexing Period: No indexing applied, the published numbers directly report quantities.
Indexing Type: No indexing applied, the published numbers directly report quantities.
Geographies: National, State, County, Metro.
Breakdowns:
Initial Claims
Continued Claims
Data Masking: No masking is performed by Opportunity Insights, but county-level data is subject to varying masking rules implemented by the state agencies that release the data. For more details, check with the relevant state agency for that state’s particular masking rules.
Notes: Unemployment claims rates are calculated by dividing unemployment claims counts by the Bureau of Labor Statistics labor force estimates from 2019.
Under the CARES Act, all states provide 13 additional weeks of federally funded Pandemic Emergency Unemployment Assistance (PEUC) benefits to people who exhaust their regular state benefits. Under the Act, through the end of 2020, some people who exhaust all these benefits, and others who have lost their jobs for reasons arising from the pandemic but who are not normally eligible for UI in their state, are eligible for Pandemic Unemployment Assistance (PUA). “Combined Claims” are defined as the sum of regular, PUA and PEUC unemployment benefit claims.
National totals for all programs’ unemployment benefit claims are the sum of the claims counts for all states and DC and exclude other territories such as Puerto Rico and the U.S. Virgin Islands.
Summary: Number of students using Zearn Math, a curriculum from the non-profit Zearn, among schools that already used Zearn Math in course instruction before the pandemic.
Data Source: Zearn
Update Frequency: Weekly, except during summer and winter school breaks.
Date Range: January 6, 2020 until the most recent date available. The data series is not updated during summer or winter school holidays.
Data Frequency: Weekly data points, with each week ending on Sunday.
Indexing Period: January 6 to February 7, 2020
Indexing Type: Change relative to the January 2020 index period, not seasonally adjusted.
Geographies: National, States, County, Metro
To ensure privacy, the data we obtain are masked such that any county with fewer than two districts, fewer than three schools, or fewer than 50 students on average using Zearn Math is excluded. Where possible, masked county levels values are replaced by commuting zone means.
Breakdowns:
School Income. Schools are classified by income based on the share of students in the school eligible for free and reduced lunch based on data provided by Zearn.
Data masking: Data is masked such that any county with fewer than two districts, fewer than three schools, or fewer than 50 students on average using Zearn Math during the period from January 6 to February 7, 2020 is excluded. Masked county level data is replaced with the commuting zone average so long as there are more than two school districts in the commuting zone or at least three schools in the commuting zone. If these condition are not met the county-level data remains masked. Additionally we exclude schools who did not have at least 5 students using Zearn Math for at least one week from January 6 to February 7, 2020.
Summary: Number of lessons completed by students each week using Zearn Math, among schools that already used Zearn Math in course instruction before the pandemic.
Data Source: Zearn
Update Frequency: Weekly, except during summer and winter school breaks.
Date Range: January 6, 2020 until the most recent date available. The data series is not updated during summer or winter school holidays.
Data Frequency: Weekly data points, with each week ending on Sunday.
Indexing Period: January 6 to February 7, 2020
Indexing Type: Change relative to the January 2020 index period, not seasonally adjusted.
Geographies: National, States, County, Metro
To ensure privacy, the data we obtain are masked such that any county with fewer than two districts, fewer than three schools, or fewer than 50 students on average using Zearn Math is excluded. Where possible, masked county levels values are replaced by commuting zone means.
Breakdowns:
School Income. Schools are classified by income based on the share of students in the school eligible for free and reduced lunch based on data provided by Zearn.
Data Masking: Data is masked such that any county with fewer than two districts, fewer than three schools, or fewer than 50 students on average using Zearn Math during the period between January 6 to February 7, 2020 is excluded. Masked county level data is replaced with the commuting zone average so long as there are more than two school districts in the commuting zone or at least three schools in the commuting zone. If these condition are not met the county-level data remains masked. Additionally we exclude schools who did not have at least 5 students using Zearn Math for at least one week from January 6 to February 7, 2020.
Summary: The daily count and rate per 100,000 people of confirmed COVID-19 cases, deaths, hospitalizations, or tests performed.
Data Source: The New York Times, The Johns Hopkins Coronavirus Resource Center, U.S. Department of Health & Human Services, Centers for Disease Control and Prevention
Update Frequency: Daily
Date Range: January 22, 2020 until the most recent date available.
Data Frequency: Daily, presented as a 7-day moving average or 7-day rolling sum
Indexing Period: No indexing applied, the published numbers directly report quantities.
Indexing Type: No indexing applied, the published numbers directly report quantities.
Geographies: National, State, Country, Metro
Breakdowns:
Data Masking: No masking is performed by Opportunity Insights.
Summary: Percentage of the population who have received one or more doses of any COVID-19 vaccine, completed a COVID-19 vaccination series, or received a COVID-19 booster or additional dose.
Data Source: The Centers for Disease Control and Prevention
Update Frequency: Daily
Date Range: February 24, 2021 until the most recent date available.
Data Frequency: Daily, presented as a 7-day moving average for new vaccinations
Indexing Period: No indexing applied, the published numbers directly report quantities.
Indexing Type: No indexing applied, the published numbers directly report quantities.
Geographies: National, State, County, Metro
Breakdowns:
Data masking: No masking is performed by Opportunity Insights.
Notes: CDC data published prior to February 24, 2021 used a different methodology to assign vaccinations to the state where they were administered, producing numbers that are not directly comparable to those published after February 24.
Summary: Time spent away from home, estimated using cellphone location data from Google users who have enabled the Location History setting.
Data Source: Google COVID-19 Community Mobility Reports, American Time Use Survey
Update Frequency: When released by Google, typically every 4-7 days.
Date Range: February 24, 2020 until the most recent date available.
Data Frequency: Daily
Indexing Period: January 3 to February 5, 2020
Indexing Type: Change relative to the January 2020 index period, not seasonally adjusted.
Geographies: National, State, County, Metro
Breakdowns:
Data Masking: Google does not release data for geographies where their internal quality and privacy thresholds are not met. Therefore some geographic areas are omitted from the series for certain breakdowns and certain dates.
Notes: When data is missing for 1 or 2 consecutive days we linearly interpolate the missing values and construct the 7 day moving average including these interpolated values. If data is missing for 3 or more consecutive days, the corresponding 7 day moving average is also recorded as missing whenever it overlaps with the missing data.
Time Away From Home is calculated by multiplying the mean time spent inside home from the American Time Use Survey by the percent change in time spent at residential locations reported by Google. For more information about this imputation, see the accompanying paper.
Summary: Key state-level policy dates relevant for changes in other series trends and values. Includes start and end of stay at home order dates, public school closure dates, and non-essential business closure and re-opening dates.
Data Source(s): New York Times, MCH Strategic Data, the Institute for Health Metrics and Evaluation, and local news and government sources.
Update Frequency: This file is not being updated with data beyond June 30, 2022.
Geographies: State