economic about Data Analysis

Demography

The scientific study of population.

– U.S. Census Bureau • Decennial Census collected every 10 years since 1790.

– Worlds largest data set.

– Determines the number of congressional representatives and allocation of federal funds.

– Census Form

• American Community Survey (ACS) sample that supplements the census with ongoing data gathering on additional topics (housing, education, occupation, etc.).

– Center for Disease Control (CDC) • Data on diseases, life expectancy, drug use, obesity, behaviors, etc.

• Records vital stats (births, deaths, marriages & divorces)

– Pew Research Organization • Various surveys on such topics as immigration, personal finance, political affiliation,

and attitudes.

Demography

Demography

Issues with Census Data:

• Self enumerations may undercount specific groups – Privacy issues, mistrust of government, and/or inability to locate may limit

participation by minorities, inner city residents, homeless, and transients.

– Reduces political representation and funding.

• Prisoners count as residents of the prison – Prisoners are disproportionally adult minority males, skewing geographical

demographics.

– May add to political representation and funding in location of prison.

• Inter-census year data are estimates only – Population changes are based on county birth and death data.

– County housing records are then used to allocate the population growth to individual cities within each county.

– Creates large gaps between decennial headcounts relative to the prior year.

Demography Issues with Census Data: • Privacy

– Data is adjusted to preserve anonymity without sacrificing demographic patterns. • Identities of respondents are removed. • Income values are rounded off. • Outliers are averaged together. • Characteristics of respondents are swapped.

Researching Undocumented Immigrants • Lowest estimates come from surveys since many are hesitant to reveal their

undocumented status out of fear of deportation. • Medium estimates come from a residual approach that involves subtracting

legal immigrants from the entire foreign-born population in the U.S. • Highest estimates come from Border Patrol extrapolations measuring arrests at

the border; however, these are biased since the same individual may be arrested multiple times.

• Accurate counts are critical! – Undocumented residents count for congressional apportionment – Allows for better cost/benefit analysis of migrants and policy prescriptions.

Demography Researching Race and Ethnicity • Non-scientific conflations of biological, national origins, and/or linguistic

traits. • Census provides multiple categories of race but no “multi-racial” category. • Who is “Black” or “African American”

– NAACP estimated that despite 70% of Blacks being multi-racial, only 3% checked more than one box. – CDC’s Vital Statistics definition historically assigned the race of the non-white parent to the child; since 1989

they have used the mother’s race (led to an increase in black infant mortality rates). – Race at death often involves a visual inspection of the body by a mortician or physician.

• Who is “Asian” – Typically identified by country of origin. – Write-in surveys are especially problematic for uneducated groups, causing an undercount.

• Who is “Hispanic” – Broader definition using cultural characteristics; has led to increased political power.

• Acquired an entirely separate question on Census form.

• Who is “Arab” or “Middle Eastern” – No separate category in census.

• Summary – Imbalances in political representation and funding for certain groups. – Death rates often use mortician/physician evaluation of race in numerator but census evaluation in

denominator. – Inconsistent results, lack of clear definition cause people to often choose different categories at different

times in their lives.

Demography Researching LGBT Community • 1948 Kinsey study contended 10% of the population is homosexual.

– Sample bias: males studied were incarcerated and included prostitutes and sex offenders.

• 1992 national opinion poll showed 2.8% (identify as gay), 6% (attracted to same sex), and 9% (had at least on homosexual experience since puberty). – Self-selection bias: volunteers may not have been representative of all consumers.

• 1993 Yankelovich Consumer Survey found 5.7% of respondents were gay. – Self-selection bias: volunteers may not have been representative of all consumers.

• 2011 Researcher Gary Gates averaged four national and two state surveys conducted after 2000 and concluded about 3.5% self identify as Lesbian, Gay, or Bisexual. – Sample bias: one of the surveys was in California (highest gay population in the U.S.)

• Summary – Sample and self-selection biases limit the credibility of many studies. – Surveys conducted in specific geographies may not be representative of the larger

population. – Personal nature implies survey method (online, phone, mail, personal interview) may yield

inconsistent results. – Phrasing: different interpretations of “Transgender”, “Bi”, “Homosexual”, “Gay”. – Sexual behavior may differ from sexual orientation and gender identity.

Demography

Researching Households • Census identifies “Household” by the housing unit, not the relationship of inhabitants.

• “Family” vs. “Non-Family” households: family is two people or more people related by birth, marriage, or adoption and residing together.

• Recent rulings on gay marriage suggest Household composition may shift from “Non- family” to “Family”.

• Many research projects analyze “family households”, omitting young affluent, single, and/or cohabiting individuals.

– May bias income, housing, education, employment, and other key statistics.

Demography

Researching Marriage and Divorce • Divorce & Marriage

– Since the 1980s divorces per 1000 people have fallen. • Stat controls for population changes but not the number of marriages.

• Over same time frame number of marriages has fallen too.

• Is the lower number of divorces because of lower marriages failing or just less marriages?

– Longitudinal studies estimate the marriage survival rate • For marriages occurring in the 1970s the 25-year rate was 48% (typical media point that half of

marriages fail)

• From 2006-2010 the survival rate for first marriages was: – 10 year: 68% for women and 78% for men.

– 20 year: 52% women and 56% men.

– Details Matter • Divorce rates are much lower for those that marry older relative to those that marry young.

• Cohabitation vs. Marriage – Decline in married households is partly due to a substitution toward long-term

cohabitation.

– In 2002 >20% of cohabitating couples had lived together for >5 years, suggesting a long-term arrangement.