Research ArticleSCIENCE POLICY

What does Congress want from the National Science Foundation? A content analysis of remarks from 1995 to 2018

See allHide authors and affiliations

Science Advances  14 Aug 2020:
Vol. 6, no. 33, eaaz6300
DOI: 10.1126/sciadv.aaz6300
  • Fig. 1 Number of discussions, by House/Congress.

    CR, Congressional Record.

  • Fig. 2 Democratic and Republican use of words, compared.

  • Fig. 3 Structural topic model.

    Based on sentence-level analysis, in (nontitle) sentences that include a direct mention of the NSF. Text is pre-whitened to remove numbers and standard stopwords, and words are lemmatized for the estimation of the STM. The words used to label each topic are shown on the left. “Highest Prob:” indicates words that occur most frequently in each topic. “FREX:” indicates words that are frequent and exclusive to each topic (16). Topics that are significantly more likely to be evident in speeches by Democrats or Republicans are shown in blue or red, respectively, where topic probabilities are shown on the right, based on coefficients and SEs for a 10-topic structural topic model with Democrat/Republican as a predictor. This visualization of results draws directly on (17).

  • Fig. 4 Sentiment in NSF sentences, by party and Congress.

  • Table 1 Most frequent words co-occurring with NSF, 1995–2017.

    Top 20 wordsCount
    Program2061
    Fund1644
    Year1394
    Education1177
    Agency970
    Technology940
    Bill896
    Act879
    Grant814
    Support765
    Department749
    Math651
    Federal636
    Million615
    Fiscal588
    Budget570
    Director546
    Engineering545
    Amend527
    Health498
  • Table 2 Most distinguishing words co-occurring with NSF, 1995–2006 to 2007–2018.

    Number in parentheses denotes ranking over the entire period 1995–2018. Findings based on sentence-level analysis, in (nontitle) sentences that include a direct mention of the NSF (see Materials and Methods for details on pre-whitening of text). Words in bold are the top five words mentioned over the entire time period. “Most distinguishing words” for each period are the top 20 diff scores, where for each word w, the following is calculated for the first time period (P) and the second time period (Q), respectively: diffwP=countwP#sentencesPcountwQsentencesQ, and diffwQ=countwQ#sentencesQcountwPsentencesP.

    The 20 words that most distinguish the time periods:
    1995–20062007–2018
    YearProgram
    FiscalBill
    BudgetTeacher
    AgencyEducation
    PercentGrant
    ExpensesFund
    NecessaryDepartment
    MillionStem
    FederalTeaching
    SecretaryStandards
    CarryingInnovation
    ProvidedNist
    CommitteeInterest
    BasicEnergy
    ActivityPartnership
    EnvironmentalOffice
    EpaInstitute
    SalaryScholarship
    MayTechnology
    DirectorBillion
  • Table 3 Most common words co-occurring with NSF, by party.

    Based on sentence-level analysis, in (nontitle) sentences that include a direct mention of the NSF (see Materials and Methods for details on pre-whitening of text). Words in bold are the top five words mentioned over the entire time period.

    Twenty most common words for…
    DemocratsRepublicans
    ProgramProgram
    FundFund
    YearYear
    EducationAgency
    BillEducation
    AgencyBill
    TechnologyGrant
    DepartmentMillion
    SupportTechnology
    GrantDepartment
    MillionAct
    MathSupport
    ActFederal
    BudgetAmend
    FederalFiscal
    NasaBudget
    EngineeringMath
    HealthHealth
    DevelopmentDirector
    BasicAdministration
  • Table 4 Most distinguishing words co-occurring with NSF, 1995–2006 to 2007–2018.

    Based on sentence-level analysis, in (nontitle) sentences that include a direct mention of the NSF (see Materials and Methods for details on pre-whitening of text). Words in bold are the top five words mentioned over the entire time period. Most distinguishing words for each period are the top 20 diff scores, where for each word w, the following is calculated for the first time period (P) and the second time period (Q), respectively: diffwP=countwP#sentencesPcountwQsentencesQ, and diffwQ=countwQ#sentencesQcountwPsentencesP

    Twenty words that distinguish between time periods:
    4D. Democrats only:4R. Republicans only:
    1995–20062007–20181995–20062007–2018
    BudgetProgramYearGrant
    EpaEducationProgramInterest
    HousingFundFiscalTaxpayer
    PlantTechnologyAgencyAccountable
    ProvidedBillMillionDollars
    HudActSupportSpending
    ResolutionDepartmentActMoney
    CommitteesGrantFederalAmerican
    FiveSupportShallChange
    AlmostMathAdministrationClimate
    CoastalStandardsBudgetInnovation
    ParticularlyTeacherEducationPolitical
    PercentEngineeringAuthorizePeople
    NsfsInstituteActivityStem
    InternetEnergyBasicFunds
    NationsShallSectionSpent
    GenomeSectionHealthMission
    HalfYearEngineeringAwarded
    IncreasedBillionFundSure
    TimeNistEnvironmentalEarlier
  • Table 5 Most distinguishing words co-occurring with NSF, 1995–2018, by party.

    Based on sentence-level analysis, in (nontitle) sentences that include a direct mention of the NSF. Text is pre-whitened to remove numbers and standard stopwords. Words in bold are the top five words mentioned over the entire time period (by both parties). Most distinguishing words for each party are the top 20 diff scores, where for each word w, the following is calculated for Democrats (D) and Republicans (R), respectively: diffwD=countwD#sentencesDcountwRsentencesR, and diffwR=countwR#sentencesRcountwDsentencesD

    Twenty most distinguishing words for…
    Democrats:Republicans:
    ProgramMoney
    FundAmend
    BillProject
    EducationExpenses
    TechnologyNecessary
    MathHuman
    NasaBoard
    StudentsProjects
    TeacherStudy
    InstituteReport
    DepartmentAdministration
    BudgetFiscal
    SupportDirector
    InnovationTaxpayer
    NistActivity
    EngineeringCommission
    InstitutionsEnsure
    DevelopmentAuthorize
    DoublingAccount
    UniversitiesCommittee

Supplementary Materials

  • Supplementary Materials

    What does Congress want from the National Science Foundation? A content analysis of remarks from 1995 to 2018

    A. Lupia, S. Soroka, A. Beatty

    Download Supplement

    This PDF file includes:

    • Tables S1 to S4

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article