Back to Dashboard

Anchor Economy Dashboard


Methodology for the Anchor Economy Dashboard.

1. Industry Definitions

For the purposes of this study, the following North American Industry Classification System (NAICS) industry codes are used to define the higher education and hospital sectors, respectively:

  • Hospitals are defined using NAICS code 622.
  • Higher education is defined using NAICS codes 6112 and 6113.

The industry codes selected are consistent with the numerous government data sources that are used throughout this study. It is worth noting that specific industries related to higher education and hospitals are not considered as part of this study, including those industries within the broader health care and social assistance sector (NAICS 62):

  • Ambulatory health care services (NAICS 621)
    • Offices of physicians
    • Offices of dentists
    • Offices of other health practitioners
  • Nursing and residential care facilities (NAICS 623)
  • Social assistance (NAICS 624)

As well as those industries within the broader educational services sector (NAICS 61):

  • Elementary and secondary schools (NAICS 6111)
  • Business schools and computer and management training (NAICS 6114)
  • Technical and trade schools (NAICS 6115)
  • Other schools and instruction (NAICS 6116)
  • Educational support services (NAICS 6117)

2. Institutional Data

As part of the 2004 Anchor Dashboard update, institutional data were compiled from the Integrated Postsecondary Education Data System (IPEDS), the Quarterly Census of Employment and Wages (QCEW), and American Hospital Association (AHA).


The following categories of IPEDS data were compiled for each region:

  • Total enrollment, broken out by graduate and undergraduate
  • Total completions
  • Number of institutions by Carnegie classification of type of institution

Because, unlike in 2019, 2004 IPEDS data do not include the county for each higher educational institution, it was necessary to geocode the addresses for institutions that did not match to a later year in which the county was provided.


American Hospital Association data on number of community hospitals, number of beds, number of admissions, and number of surgical operations for the year 2004 were purchased by the Federal Reserve Bank of Philadelphia and provided to Oxford Economics and then collated by region. Note that some MSAs in our data set were not included in the AHA data. Data for these MSAs are, therefore, missing.


The number of establishments in higher education (NAICS 6112 and 6113) and in hospitals (NAICS 622) are reported based on QCEW establishment counts. This sums the number of institutions regardless of ownership (i.e., including both private and local, state, and national government–owned establishments), which corresponds to the establishments whose activity is modeled in the economic modeling work described in section 3. Establishment counts are based on county-level data, which are summed to obtain regional totals for MSA and non-MSA regions.1 Establishment counts are used in the modeling for economic impact but are not reported separately in the Anchor Economy Dashboard, as they include offices and locations that may be purely administrative in nature and not patient- or student-facing.

In California, institution counts from the QCEW for both hospitals and colleges proved unreliable. For this reason, for California regions, the higher education establishment count is derived from the count of IPEDS postsecondary educational institutions, and the hospital establishment count is derived from the count of AHA community hospitals.

Note that the QCEW is only used as the direct source for the establishment count variables (outside of California). We note that direct employment and labor income impacts are analogous to the employment and wage variables in the QCEW,2 summed across geography (counties) and ownership (private/government) categories. (That is, direct employment represents workers at the institutions themselves, which is what is reported in QCEW.) However, the original QCEW data have a large number of missing values for these variables because of data suppressions by the Census. (The establishment counts described above are not suppressed.) IMPLAN uses these QCEW values as a major input into its estimates of employment and labor income by county and industry, along with other government data sources, and so effectively estimates values for suppressed data in the QCEW. We thus use the IMPLAN estimates for the four variables of higher ed employment, higher ed income, hospital employment and hospital income, rather than the raw QCEW data with suppressions. This approach matches that of the 2019 work.

3. Economic Impact Modeling

Following the methods used in 2019, three sources of economic impacts are calculated in the economic impact modeling:

  • The operational activity (operational spend) of hospitals and colleges. This includes the compensation of employees, capital income to business owners, certain directly paid taxes (taxes on production and imports), and the full impact from supply chain spending. (Section 3.1)
  • Ancillary retail spending associated with visitors to hospitals and colleges. (Section 3.2)
  • Impacts from capital expenditures (capital spend) associated with hospitals and colleges. (Section 3.3)

The inputs were run through the IMPLAN system for the 524 regions to calculate, for each of the three types of impacts, three channels of impact. These three channels of impact are standard for input-output economic impact analysis and are represented in Figure 1.

  • Direct impacts, representing the economic activity at the institutions themselves;
  • Indirect impacts, representing the full supply chain impact; and
  • Induced impacts, representing the economic activity supported by spending of direct and indirect workers out of wages.

Note that each of the three sources of impact has associated with it the three channels of impact, and that these can also be disaggregated between eds and meds. Thus, one might consider the indirect impact from ancillary spending at educational institutions, or the induced impact of capital expenditures at hospitals, for example. Each of these is measured in terms of employment, GDP, and labor income. Section 4 discusses how these various results are presented in the dashboard.

Figure 1. Summary of the Channels of Economic Impact

Direct effects: The first group of impacts to be assessed is the economic activity associated with the eds and meds U.S. operations. This is defined as the activity supported by the direct employment and sales of the industries.

Indirect (supply chain) effects: This type of impact identifies linkages between eds and meds and the sectors’ respective supply chains. As a result of purchasing goods and services from suppliers, economic value is created beyond the direct operations of the sector. This includes, for example, jobs supported in a wide variety of activity in publishing, medical equipment manufacturing, and business services sectors (IT, accounting, auditing, etc.). Of critical importance when estimating multipliers is to consider leakage. This concept captures the fact that some purchases will be made outside the region (or even country) and do not add to regional output or employment.

Induced (workers’ spending) effects: The induced impact captures economic activity supported by those directly or indirectly employed by eds and meds who spend their disposable income on goods and services in the regional economy. This helps support jobs in the industries that supply goods and services to consumers, including jobs in retail outlets, restaurants, and a range of other service industries. This is also estimated in terms of regional gross domestic product (GDP) and employment.

How the Channels of Impact Are Measured

The channels of impact (direct, indirect, and induced) are quantified across three primary measures: employment, income, and gross domestic product (GDP) . Each category is defined below, and the impacts are calculated across each of the aforementioned channels for each of the 546 IMPLAN industries. The results of the contribution analysis are presented as the summed total of all 546 industries.

Employment: An industry-specific mix of full-time, part-time, and seasonal employment. An annual average that accounts for seasonality and follows the same definition used by the Bureau of Labor Statistics (BLS) and the Bureau of Economic Analysis (BEA). IMPLAN employment is not equal to full-time equivalents.

Income (Labor income): All forms of employment income, including employee compensation (wages and benefits) and proprietor income.

Gross Domestic Product (GDP): A measure of output-less intermediate consumption that represents an industry’s contribution to GDP. It is the measure of the value of goods and services produced in a specified region. The previous iteration of the Anchor Economy Dashboard referred to this same measure as gross value added, which is a slightly narrower measure of economic output that excludes certain direct taxes. Because IMPLAN includes these relevant taxes, we opted to use gross domestic product in this iteration, given its familiarity.

3.1 Operational Spend (“opex”)

Operational impacts are calculated for both of the two industry categories (eds and meds) and for four ownership categories (private and local, state, and federal governments). This results in seven types of institutions included (the omitted category being federal government–owned colleges, which do not exist — at least in the economic data).

For each of these seven, the scale of the impact is determined as a share of a relevant IMPLAN industry, except the federal government–owned hospitals, for which QCEW data are used instead. The advantage of using IMPLAN data, which are based on a number of different underlying government data sets, including the QCEW, is that IMPLAN has already performed key data cleaning tasks such as estimating values that are suppressed for data privacy reasons in the underlying government data, and that it’s consistent with the rest of the data used in the economic impact estimates.

The table below presents the sources for each of the seven categories. For privately owned colleges and hospitals, the share of the underlying IMPLAN industries used in the impact analysis is 100 percent. For state and local government–owned hospitals and colleges, the share of the IMPLAN industry included in the impacts is estimated using data from the Annual Survey of State and Local Government Finances (ASSLGF). For education, this share is the percentage of total state or local educational spending that goes to higher education institutions (as opposed to primary and secondary schools). For hospitals, the share is the percentage of state or local health-care spending that goes to hospitals (as opposed to other types of public health spending). For federal government–owned hospitals, the impacts are estimated based on the employment and employee compensation at federally owned hospitals in the QCEW data.3

Eds/meds Ownership IMPLAN
IMPLAN industry title Impact amount
Eds Private 481 College & universities 100% of IMPLAN ind.
Eds State government 539 State government, education Share of IMPLAN ind. based on ASSLGF
Eds Local government 542 Local government, education Share of IMPLAN ind. based on ASSLGF
Meds Private 490 Hospitals 100% of IMPLAN ind.
Meds State government 540 State government, hospitals and health services Share of IMPLAN ind. based on ASSLGF
Meds Local government 543 Local government, hospitals and health services Share of IMPLAN ind. based on ASSLGF
Meds Federal government 546 Federal government, non-military Employment & employee compensation of federally-owned hospitals in QCEW

Special Inputs for State Schools in Six MSAs

The standard input for state government–owned colleges, described above, was overridden for six specific MSAs in which the IMPLAN imputation of suppressed QCEW employment in this industry and ownership category appears to have been of poor quality. In these six MSAs, data from IPEDS are used instead to estimate the employment for state government–owned colleges. These six MSAs are:

Region State College
State College, PA Penn State University
Corvallis, OR Oregon State University
Lawrence, KS University of Kansas
Champaign-Urbana, IL University of Illinois
Blacksburg-Christiansburg-Radford, VA Virginia Tech (State University)
Harrisonburg, VA James Madison University

3.2 Ancillary Spending Impacts

Impacts are also estimated for retail spending associated with visitors to colleges and hospitals. Three types of spending are estimated: conference spending at both hospitals and colleges, and hospital visitor spend.4 Note that these impacts only consider the retail sales margins associated with this spending, not the impacts of producing the goods that are purchased.

Eds/meds Type of spend Unit
spend (2019)
Scaling factor Scaling
factor source
Eds Conference spend $14.56 CEX # research & professional staff IPEDS
Meds Conference spend $14.56 CEX Employment in healthcare occupations OEWS
Meds Hospital visitor spend $4.25 CEX # of surgeries AHA

The amount of spend per conference attendee and per hospital visitor was estimated for 2019 based on data from the Consumer Expenditure Survey (CEX) at $14.56 and $4.25, respectively. For the 2004 work, we inflation-adjust these values to 2004 dollars. The number of individuals to apply these per person spend values to is sourced for educational conference spending from the number of research and professional staff in IPEDS, for health-care conferences on the regional employment in health-care occupations in the Occupational Employment and Wages Survey (OEWS),5 and for hospital visitor spending from the number of surgeries performed annually based on data from the AHA.

3.3 Capital Spending Impacts (“capex”)

Capital spending impacts are estimated separately in a single national model and distributed to the regions based on the direct employment impacts in the relevant category (i.e., separately for eds and for meds) from the IMPLAN operational modeling described in section 3.1.

The source for the capital spending assumptions is the BEA Investment in Fixed Assets data by industry.6 Investment data for the education sector (6100) and for hospitals (622H) were aligned to IMPLAN categories to estimate the national-level impact. Because educational investment is not available separately for institutions of higher education, the capital expenditure for this sector was scaled to (private) higher educational institutions’ share of this sector based on IMPLAN data (53 percent). The impact results from the national capex model were then distributed between regions based on their share of the national direct employment impacts from the operational modeling. Because the BEA capex expenditures only cover privately owned institutions, additional capex impact was allocated for government-owned institutions using the same relationship between direct employment and capital impacts derived for private institutions.

4. Location Quotients, Reliance Scores, and Multipliers

IMPLAN data for each of the 524 regions were used to calculate the location quotients (LQs) for the employment, GDP, and labor income impacts. LQs are defined as the ratio of the relevant impact to the total employment, GDP, or labor income for the relevant region divided by the analogous ratio for the United States as a whole. For example, if the total employment impact in a particular region represents 6 percent of employment in the region, and nationally the employment impact is 4 percent of national employment, the employment LQ would be 1.5 (= 6% / 4%).

The reliance score is defined for this work as the average of the three LQs.

Multipliers for employment and labor income are also provided and are defined as the ratio of the total impact (measured in employment or labor income) divided by the direct impact. In the original version of the 2019 Anchor Economy Dashboard, the multipliers were defined instead as the ratio of the total impact to the direct impact only for the opex activity — i.e., excluding capex and ancillary impacts in the denominator. These multipliers have been adjusted in the 2019 data set, and this change is noted again below in “Adjustments to the 2019 Results.”

5. Dollar Year Adjustments

The base results for the 2004 analysis are in 2004 dollars. All dollar amounts displayed in the dashboard data have been inflation-adjusted to 2019 dollars. To aid comparison with the 2019 work, results are adjusted to 2019 dollars using industry-specific adjustment factors provided by IMPLAN. Dollar values for GDP and labor income were adjusted in the original 2004 detailed modeling results for each of the 546 IMPLAN industries. Regional data were also adjusted, and these adjusted regional data were used to calculate LQs and reliance scores for 2004 as well.

6. Adjustments to the 2019 Results

The following adjustments have been made to the 2019 data to make them fully comparable with the new 2004 data:

  • The employment and labor income multipliers (multiplier_emp, multiplier_inc), which were previously calculated as the ratio of the total impact (across operational, capital, and ancillary expenditures) to the direct operational expenditure impact only are now the ratio of the total impact to the direct impact across operational, capital, and ancillary activities.
  • The two variables highered_inc_qcew and hospital_inc_qcew were previously the direct operational employee compensation; they are now the direct operational labor income. The distinction between employee compensation and labor income is that the latter includes self-employment income by self-employed workers, whereas the former does not. Using labor income instead of employee compensation is a small change that makes these variables more consistent with the other variables in the data set.
  • The variables highered_opex, highered_capex, highered_ancillary, hospital_opex, hospital_capex, and hospital_ancillary in the previous data are now defined as the direct GDP impact in eds and meds for opex, capex, and ancillary activities. Thus, they sum together to the overall direct GDP impact. See the table below for updated variable names.
  • The variables highered_estab_qcew and hospital_estab_qcew have been removed, given that the QCEW establishment counts may include offices and locations that may be purely administrative in nature and not patient- or student-facing, and thus may be misleading.

Given discrepancies in data sources between 2004 and 2019, variables conveying institution characteristics (e.g., enrollment, degrees conferred, hospital admissions) were removed from the downloadable data, since these same fields could not be replicated for 2004.

7. State-Level Anchor Impacts

State-level anchor impacts are available on the Anchor Economy Dashboard for each state and the District of Columbia for both 2004 and 2019. Impacts for states where all regions fall within the state’s geography are aggregated to retrieve state totals. These totals are then used to calculate the location quotients for anchor employment, income, and GDP, which we average together to find the reliance index. For states where regions fall within the boundaries of multiple states (for example, St. Louis MO-IL), regional impacts were disaggregated to the state level by weighting the multistate regional impacts by the share of a region’s population that lived in the state. After calculating weighted impacts for the share of each region within each state, we aggregated the total anchor impacts of all regions in each state to be used in the location quotient and reliance index calculations.

8. Variable Description and Source

This table outlines all metrics provided in the Anchor Economy Dashboard, including the channel of impact they represent (direct, indirect, induced, or total) and the measure they capture (income, employment, or GDP). In the downloadable data, variables for 2004 and 2019 are divided into separate worksheets.

9. U.S. Regions Reported in the Anchor Economy Dashboard

Data and the reliance index in the Anchor Economy Dashboard are at the metro/nonmetro regional level as defined by the BLS. A definition of regions can be found here. Geographic boundaries remain consistent across the 2004 and 2019 data.


The QCEW also provides direct totals for privately owned establishments by MSA; however, because we are interested in government-owned establishments as well as non-MSA regions, it was necessary to sum the county-level data. Note that the county-summed establishment count for privately owned establishments does not perfectly match the MSA-level totals but is typically close (within one or two establishments).


Labor income from IMPLAN, however, is more inclusive than wages in the QCEW, and it includes both the value of nonwage compensation like benefits as well as self-employment income.


QCEW data for federally owned institutions are not suppressed, as they are for privately owned and state and local government–owned institutions when there are data privacy concerns.


A fourth spending type, associated with spending by out-of-state students, was ultimately excluded from the 2019 work and therefore is also excluded from the 2004 update.


The OEWS was chosen because it largely aligns to the geographies used in the study and because it allows the identification of health-care workers by occupation.


See the nonresidential detailed estimates for investment: