| Title: | Steve's Toy Data for Teaching About a Variety of Methodological, Social, and Political Topics |
|---|---|
| Description: | This is a collection of various kinds of data with broad uses for teaching. My students, and academics like me who teach the same topics I teach, should find this useful if their teaching workflow is also built around the R programming language. The applications are multiple but mostly cluster on topics of statistical methodology, international relations, and political economy. |
| Authors: | Steve Miller [aut, cre]
|
| Maintainer: | Steve Miller <[email protected]> |
| License: | GPL-2 |
| Version: | 1.4.0 |
| Built: | 2024-10-27 12:33:50 UTC |
| Source: | CRAN |
These data are in Table 9.1 of the 3rd edition of Agresti and Finlay's Statistical Methods for the Social Sciences. The data are from Statistical Abstract of the United States and most variables were measured in 1993.
af_crime93af_crime93
A data frame with 51 observations on the following 8 variables.
statea character vector for the state
violenta numeric vector for the violent crime rate (per 100,000 people in population)
murdera numeric vector for the murder rate (per 100,000 people in population)
povertya numeric vector for the percent with income below the poverty level
singlea numeric vector for the percent of families headed by a single parent
metroa numeric vector for the percent of population in metropolitan areas
whitea numeric vector for the percentage of the state that is white
highschoola numeric vector for the percent of state that graduated from high school
The data are from Statistical Abstract of the United States and most variables were measured in 1993. These data should result in regressions that would flunk a Breusch-Pagan test for heteroskedasticity.
Agresti, Alan and Barbara Finley. 1997. Statistical Methods for the Social Sciences. Prentice Hall. (3rd Edition)
A data set on modeling coups in Africa using data from the period between 1960 and 1975 (1982). These data offer a partial replication of Jackman (1978).
african_coupsafrican_coups
A data frame with the following 11 variables.
iso3ca three-character ISO code for state identification
countryan English country name
jciJackman's (1978) coup index from 1960 to 1975
tmisJohnson et al.'s (1984) total military involvement score
agpercan estimate of the percentage of the country's labor force in agriculture and animal husbandry
indpercan estimate of the percentage of the country's labor force in industry
literacy_cntsan estimate of countrywide literacy from around 1965
literacy_baanother estimate of countrywide literacy from around 1965
lepercan estimate of the size of the largest ethnic group, as a percentage
partydomthe percentage of the vote received by the largest party in the country prior to independence
turnoutthe turnout (as a percentage) at the independence referendum
Data exist for instructional purposes, especially about modeling interactions. Reading Jackman (1978) and Johnson et al. (1984) will provide more information about the data and hypotheses. There was a follow-up symposium on this in 1986 in American Political Science Review that may be an interesting read and provide even more context about what's at stake in this debate.
English country names are country names from around the time of publication.
Take note of older names of "Dahomey", "Swaziland", "Upper Volta", and "Zaire."
The three-character ISO codes are current, mostly for ease of doing other
things with the data. However, this comes with the acknowledgment that
Dahomey and Zaire used to have different ISO codes under their older names.
Both codes for Dahomey (DHY) and Zaire (ZAR) were retired in 1977 and
1997, respectively.
Ideally, I'd have Morrison's (1972) Black Africa, but I do not. I have a copy of a 1989 update, though. That's what I consulted in constructing this data set.
Jackman (1978) is deceptively opaque on what he's doing for the ethnic group variable and arguably misleads on what his turnout variable is actually from. In the case of the ethnic group variable, it's the difference between saying the largest ethnic group in Rwanda is 98% of the population versus 80% of the population. In short, it's the difference of saying whether there are any Tutsi at all in the country. Basically, I'm uncertain with what he's doing with what Morrison et al. (1989) define as "ethnic clusters".
Related: the agricultural variable is a midway point between columns B and columns C in Table 3.11 of Morrison et al. (1989). I do not think this is too far removed from what Jackman was looking at in an older version of the same data, but there will be slight differences. It's the difference of "these variables came from 1966" versus "this is an interpolation of 1960 to 1970". The latter is what I offer here. I can only do so much.
To continue this theme of the opacity in trying to reconstruct the data, Jackman (1978, p. 1265) says his social mobilization index incorporates the percentage of the labor force that is not employed in agriculture. The summary statistics he provides in fn. 4 on p. 1265 are consistent with this, at least (for the most part) in this reconstruction of the data. However, other statistical results and other language from Jackman are consistent with him using the percentage of the labor force that is employed in industry. This is not a trivial distinction either. Using the percentage of the country's labor force in industry would, in a literal sense, not strictly be "the simple sum of the percentage of the labor force in nonagricultural occupations". It would exclude those working in service industries. The data provide the opportunity to use either the industrial percentage variable or to manually create a non-agricultural labor force percentage variable as the difference between 100 and the agricultural labor force percentage variable. It makes the most sense to do the latter. The industry percentage variable comes from Table 3.14 in Morrison et al. (1989) and is likewise a midway point between 1960 and 1970.
Mercifully, the coup variables come from a replication by Johnson et al. (1984). Based on Morrison et al.'s (1989) updated data, it's not clear how Jackman could've derived some of these estimates using the formula he said he used. For example, Benin should have a score of 39 based on the information in Table 2.10 (p. 373 in Morrison et al. (1989)). Cameroon should have a 1 and not a 2, per Table 5.10 (p. 399). My comments here work under the assumption that Morrison et al. are adding information and not revising information in the second edition of the Black Africa handbook.
To be more precise, both the Jackman coup index and total military involvement variables are directly copied from Table 2 in Johnson et al. (1984) on p. 627. Missingness in the Jackman coup index variable communicates the country was not included in his original study, but was included in the Johnson et al. replication.
The literacy variables have suffices communicating where I obtained them. The Cross-National Time Series Database has a variable effectively communicating this information that I was using first in trying to recreate these data. These data come from 1965 in that data set. Jackman and Johnson et al. are assuredly using Morrison's almanac. That information is in Table 4.11 of Morrison et al. (1989), though it's possible the estimates contained therein are slightly different than what was reported in the first edition. I cannot know for sure.
Ethiopia is conspicuously missing in the party dominance variable. That's by design, and apparently its omission warranted ample discussion both by Jackman (1978) and Johnson et al. (1984). Johnson et al. (1984, fns. 4,5) argue it's a curious choice that can situationally influence the results that Jackman reports, but there are also lots of other choices made by Jackman (1978) that can influence these results.
I am 99.9% sure the turnout variable is Table 5.9 in Morrison et al. (1989). Jackman (1978) says this is from before independence but I'm confident he meant it was the turnout at the independence referendum.
Jackman, Robert W. 1978. "The Predictability of Coups d'etat: A Model with African Data." American Political Science Review 72(4): 1262-75.
Jackman, Robert W., Rosemary H. T. O'Kane, Thomas H. Johnson, Pat McGowan, and Robert O. Slater. 1986. "Explaining African Coups d'Etat." American Political Science Review 80(1): 225-49.
Johnson, Thomas H., Robert O. Slater, and Pat McGowan. 1984. "Explaining African Military Coups d'Etat, 1960-1982." American Political Science Review 78(3): 622-40.
Morrison, Donald George, Robert Cameron Mitchell, and John Naber Paden. 1989. Black Africa: A Comparative Handbook (2nd ed.). New York, NY: The Free Press.
A data set on thermometer ratings for the Democratic party, Republican party, "both major parties", and a major party thermometer index from the American National Election Studies (1978-2012).
anes_partythermsanes_partytherms
A data frame with 33830 observations on the following 19 variables.
yearthe survey year
uida unique identifier for each respondent, taken directly from the time-series files for potential merging
stateabbthe two-character abbreviation for the state of residence for the respondent
therm_demthe respondent's thermometer rating of the Democratic party
therm_gopthe respondent's thermometer rating of the Republican party
therm_bmpthe respondent's thermometer rating of "both major parties"
mptithe "major party thermometer index" score for the respondent. See details for more.
agethe age of the respondent
educatthe education-level of the respondent. 1 = 8 grades or less. 2 = high school, no diploma. 3 = high school diploma. 4 = high school "plus non-academic training". 5 = Some college, no degree (includes AA holders). 6 = BA-level degree. 7 = advanced degree, including Bachelor of Laws degrees.
urbanism1 = central cities. 2 = suburban areas. 3 = rural/small towns/outlying areas.
pid71 = Strong Democrat. 2 = Weak Democrat. 3 = Independent, lean Democrat. 4 = Independent. 5 = Independent, lean Republican. 6 = Weak Republican. 7 = Strong Republican
incomepercrespondent's household income percentile. 1 = 0-16 percentile. 2 = 17-33. 3 = 34-67. 4 = 68-95. 5 = 96-100.
race4respondent's race-ethnicity summary. 1 = White, non-hispanic. 2 = Black, non-hispanic. 3 = Hispanic. 4 = Other.
unemployeda binary numeric vector for if the respondent is temporarily unemployed.
polintthe respondent's self-reported interest in public affairs. 1 = Hardly at all. 2 = Only now and then. 3 = Some of the time. 4 = Most of the time.
distrust_govtthe respondent's self-reported (dis)trust in the federal government's ability to do what's right. 1 = Just about always (trust the government). 2 = Most of the time. 3 = Some of the time. 4 = None of the time/never.
govt_crookedthe respondent's assessment of how many government officials are crooked. 1 = Hardly any. 2 = Not many. 3 = Quite a few; quite a lot.
govt_wastethe respondent's assessment of how much the government wastes in tax money. 1 = Not very much. 2 = Some. 3 = A lot.
govt_bigintereststhe respondent's assessment of whether the government is run by a few big interests. 0 = Run for the benefit of all people. 1 = Run by a few big interests.
The major party thermometer index is calculated as the thermometer rating for the Democratic party minus the thermometer rating for the Republican party. 100 is then added to that difference, which is then divided by 2. Fractional results are rounded to the next highest integer. Also note the coding of the "government distrust" measures. These are reverse-coded from their original scales.
Data come from ANES's time series file.
A simple data set for in-class illustration about how to estimate and interpret interactive relationships. The data here are deliberately minimal for that end.
anes_prochoiceanes_prochoice
A data frame with 5914 observations on the following 14 variables.
versionversion identifier from ANES
caseidtime-series case identifier from ANES
healthoppose/"NFNO"/favor abortion if pregnancy would hurt woman
fataloppose/"NFNO"/favor abortion if pregnancy would cause woman to die
incestoppose/"NFNO"/favor abortion if pregnancy was caused by incest
rapeoppose/"NFNO"/favor abortion if pregnancy was caused by rape
bdoppose/"NFNO"/favor abortion if fetus would be born with serious birth defect
finoppose/"NFNO"/favor abortion if having child would impose financial hardship
sexoppose/"NFNO"/favor abortion if the child will not be the sex the woman wants
choiceoppose/"NFNO"/favor abortion if woman chooses to have one
pidrespondent's partisanship (Democrat, Independent, Republican)
knowspeakerwas the respondent able to correctly identify the Speaker of the House (John Boehner)
addchoicean additive scale of the abortion scores
lchoicea continuous latent scale of pro-choice scores (from a simple graded response model)
"NFNO" = "Neither Favor Nor Oppose". All abortion prompts are on a 0-2 scale where 0 is oppose, 1 is "NFNO", and 2 is favor. The respondent's party identification is on a similar scale where 0 = "Democrat", 1 = "Independent", and 2 = "Republican". The additive scale of abortion scores has a minimum of 0 and a maximum of 16.
Data come from ANES's (2012) time series.
This is a simple data set for estimating a simple model on voter turnout from the 1984 American National Election Studies (ANES) 1984 time-series.
anes_vote84anes_vote84
A data frame with 2257 observations on the following 9 variables.
uida unique identifier for the respondent
stateabbthe state where the respondent lives (as an abbreviation)
votewhether the respondent voted (1 = yes; 0 = no)
agethe age of the respondent
educthe education-level of the respondent. See details section for more.
femalewhether the respondent is a woman (1 = female; 0 = male)
southdoes the respondent live in the south (1 = yes; 0 = no)
polintthe political interest of the respondent in the campaigns (-1 = not much interested; 0 = somewhat interested; 1 = very much interested)
govracedid the respondent's state have a gubernatorial election that same November (1 = yes; 0 = no)
The vote variable is deliberately coded where those with a value of 1 are respondents
who said they voted and the ANES was able to confirm that with voter registration records. There are purportedly
85 responses in this raw variable where the respondent said they voted, but this could not be confirmed from
registration records. Those cases are recorded as NA. The educ variable ranges from 1 (finished 8th
grade or less than that) to 10 (respondent holds an advanced degree). The uid variable is a simple sequence variable
ranging from 1 to 2257 and is calculated on the original 1984 time-series study (May 3, 1999 version)
before other recoding was done. This should allow some reproducibility for an interested user.
Data come from ANES's (1984) time series.
Daily data on the NYSE Arca Steel Index. These data are useful for me in teaching how Trump's 2018 steel tariffs didn't do much good for the steel industry.
ArcaArca
A data frame with 966 observations on the following 6 variables.
datethe date
closethe closing price
openthe opening price
highthe daily high in that day's trading
lowthe daily low in that day's trading
These data are taken from investing.com.
See: https://www.investing.com/indices/arca-steel-historical-data
This data set from Connelly et al. (2017) measures the Arctic sea ice extent in 10^6 square kilometers. It includes lower bounds and upper bounds on annual averages.
arcticseaicearcticseaice
A data frame with 115 observations on the following 4 variables.
yearthe year
valuethe annual Arctic sea ice extent (in 10^6 sq km)
ubThe upper bound of the value, provided by Connelly et al.
lbThe lower bound of the value, provided by Connelly et al.
This is for illustration of climate change for my intro students. Connelly et al. (2017) are in part a methodological paper. The data I present here are from the "rescaled (unadjusted T)" data in the second sheet from their replication files.
Connolly et al. (2017), ”Re-calibration of Arctic sea ice extent datasets using Arctic surface air temperature records”. Hydrological Sciences Journal 62(8): 1317–40.
Simple mean tariff rate for Argentina, starting in 1980. The goal is to keep these data current.
arg_tariffarg_tariff
A data frame with three variables:
countrycountry name (Argentina)
yearthe year
tariffratethe simple mean tariff rate for Argentina on all products (as a percentage)
Data come from various sources. World Bank estimates are used for 1980-1984 and 2010-2018, but see also Lora's (2012) report for the Inter-American Development Bank. The 1980-1984 estimates are actually means for 1980-1 and 1982-4 via Laird and Nogues' (1989) article in the World Bank Economic Review.
These are yearly counts on air accidents and fatalities, including measures for corporate jet accidents and hijackings. The hijackings are of particular interest to me, at least from a historical terrorism perspective.
asn_statsasn_stats
A data frame with 78 observations on the following 7 variables.
yearnumeric vector for the year
airacca numeric vector for the number of airliner accidents
airfatala numeric vector for the number of fatalities from airliner accidents
corpjetacca numeric vector for the number of corporate jet accidents
corpjetfatala numeric vector for the number of fatalities from corporate jet accidents
hijacka numeric vector for the number of hijackings/skyjackings
hijackfatala numeric vector for the number of fatalities from hijackings/skyjackings
All fatality estimates exclude ground fatalities. All accidents are hull-loss accidents. The airliner figures are for those flights with at least 14 passengers. Check https://asn.flightsafety.org/statistics/period/stats.php?cat=H2 for more.
Aviation Safety Network, a service provided by the Flight Safety Foundation.
This is the replication data for "Randomization Inference in the Regression Discontinuity Design: An Application to Party Advantages in the U.S. Senate", published in 2015 in Journal of Causal Inference. I use these data to teach about regression discontinuity designs.
CFT15CFT15
A data frame with 1390 observations on the following 9 variables.
statea numeric vector for the state. This is ultimately a categorical variable.
yeara numeric vector for the year of the election.
votea numeric vector for the Democratic vote share in the next election (i.e. six years later).
margina numeric vector for the Democratic party's margin of victory in the statewide election. This is the running variable, in RDD parlance.
classa numeric vector for the class to which each Senate seat belongs.
termshousea numeric vector for the Democratic candidate's cumulative number of terms previously served in the U.S. House.
termssenatea numeric vector for the Democratic candidate's cumulative number of terms previously served in the U.S. Senate.
populationa numeric vector for the population of the Senate seat's state.
treatmenta numeric vector that is 1 if margin > 0 and is 0 if margin < 0.
Cattaneo, Matias D. and Brigham R. Frandsen and Rocio Titiunik. 2015. "Randomization Inference in the Regression Discontinuity Design: An Application to Party Advantages in the U.S. Senate". Journal of Causal Inference 3(1): 1–24.
Cattaneo, Matias D. and Brigham R. Frandsen and Rocio Titiunik. 2015. "Randomization Inference in the Regression Discontinuity Design: An Application to Party Advantages in the U.S. Senate". Journal of Causal Inference 3(1): 1–24.
Calonico, Sebastian and Matias D. Cattaneo and Max H. Farrell and Rocio Titiunik. 2017. "rdrobust: Software for regression-discontinuity designs". The Stata Journal 17(2):372–404.
A data set on voting intentions in the 1988 Chilean plebiscite, which ultimately ended the military junta rule of Augusto Pinochet.
chile88chile88
A data frame with 2700 observations on the following 8 variables.
regiona character vector for the region of Chile in which the respondent lives
popthe population size of the respondent's community
sexa numeric vector that equals 1 if the respondent is a woman
agea numeric vector for the age of the respondent
educa character vector indicating whether the respondent has a primary (P), secondary (S), or post-secondary (PS) education
incomea numeric vector for respondent's monthly income (in pesos)
sqa numeric vector for the scale of support for the status quo in Chile
votea character vector for the vote intention of the respondent (see details)
Data were manually downloaded from John Fox's website. You will see
his version of these data as Chile in the carData package. I
changed a few things that are ultimately cosmetic. It's basically this data
set.
The vote variable communicates vote intentions, whether to vote "Yes" (Y) to continue the Pinochet regime, to vote "No" (N) to end the Pinochet regime, to abstain (A) from a vote, or whether the respondent is undecided (U). 168 respondents did not answer the question.
Fox (2008, 336) does not say much about the status quo variable, and on what scale it is. It can only be easily inferred that higher values = more support for the status quo.
You may find it in your interest to relabel the "region" variable. In these data, the regions are Central ("C"), Metropolitan Santiago area ("M"), North ("N"), South ("S"), and the city of Santiago ("SA").
More information about the underlying source of the data would be more than welcome. Any information about these data, beyond the kind of R documentation files about its pedagogical use, is hard to find. This is a roundabout way of saying be cautious about any "real-world" use of these data beyond learning statistical methods. That is ultimately its intended use.
Fox, John. 2008. Applied Regression Analysis and Generalized Linear Models (2nd ed.). Los Angeles, CA: Sage.
This data set contains daily temperatures (highs and lows) for Clemson, South Carolina from Jan. 1, 1930 to the end of the most recent calendar year. The goal is to update this periodically with new data for as long as I live in this town.
clemson_tempsclemson_temps
A data frame with 33,148 observations on the following 3 variables.
datethe date
tminthe daily low, adjusted to Fahrenheit
tmaxthe daily high, adjusted to Fahrenheit
Data obtained from NOAA, via the rnoaa package. The station identifier is GHCND:USC00381770 for added context.
The call from rnoaa returns these values initially as Celsius*10. I don't know why NOAA does it this way, but there you go.
This is a sample data set, cobbled from various sources, about carbon dioxide emissions in the history of the planet from 800,000 BCE to the most recently concluded calendar year. I use this for a data visualization example for a lecture on climate change and international politics. Data communicate yearly averages/estimates.
co2emissionsco2emissions
A data frame with 3,099 observations on the following 2 variables.
yearthe year (negative values = BCE)
valueestimated carbon dioxide emissions (in ppm)
The data come from many sources. Before 0 CE, the data come from 10 sources described by the Environmental Protection Agency ("Climate Change Indicators: Atmospheric Concentrations of Greenhouse Gases"). Observations from 0 CE to 2014 come from Meinshausen et al. (2017) doi:10.5194/gmd-10-2057-2017. Observations from 2015 forward come from NASA ("Vital Signs").
EPICA Dome C and Vostok Station, Antarctica: approximately 796,562 BCE to 1813 CE Lüthi, D., M. Le Floch, B. Bereiter, T. Blunier, J.-M. Barnola, U. Siegenthaler, D. Raynaud, J. Jouzel, H. Fischer, K. Kawamura, and T.F. Stocker. 2008. High-resolution carbon dioxide concentration record 650,000–800,000 years before present. Nature 453:379–382.
Law Dome, Antarctica, 75-year smoothed: approximately 1010 CE to 1975 CE Etheridge, D.M., L.P. Steele, R.L. Langenfelds, R.J. Francey, J.-M. Barnola, and V.I. Morgan. 1998. Historical CO2 records from the Law Dome DE08, DE08-2, and DSS ice cores. In: Trends: A compendium of data on global change. Oak Ridge, TN: U.S. Department of Energy.
Siple Station, Antarctica: approximately 1744 CE to 1953 CE Neftel, A., H. Friedli, E. Moor, H. Lötscher, H. Oeschger, U. Siegenthaler, and B. Stauffer. 1994. Historical carbon dioxide record from the Siple Station ice core. In: Trends: A compendium of data on global change. Oak Ridge, TN: U.S. Department of Energy.
Mauna Loa, Hawaii: 1959 CE to 2015 CE NOAA (National Oceanic and Atmospheric Administration). 2016. Annual mean carbon dioxide concentrations for Mauna Loa, Hawaii.
Barrow, Alaska: 1974 CE to 2014 CE Cape Matatula, American Samoa: 1976 CE to 2014 CE South Pole, Antarctica: 1976 CE to 2014 CE NOAA (National Oceanic and Atmospheric Administration). 2016. Monthly mean carbon dioxide concentrations for Barrow, Alaska; Cape Matatula, American Samoa; and the South Pole.
Cape Grim, Australia: 1992 CE to 2006 CE Shetland Islands, Scotland: 1993 CE to 2002 CE Steele, L.P., P.B. Krummel, and R.L. Langenfelds. 2007. Atmospheric CO2 concentrations (ppmv) derived from flask air samples collected at Cape Grim, Australia, and Shetland Islands, Scotland. Commonwealth Scientific and Industrial Research Organisation.
Lampedusa Island, Italy: 1993 CE to 2000 CE Chamard, P., L. Ciattaglia, A. di Sarra, and F. Monteleone. 2001. Atmospheric carbon dioxide record from flask measurements at Lampedusa Island. In: Trends: A compendium of data on global change. Oak Ridge, TN: U.S. Department of Energy.
Meinshausen, M., Vogel, E., Nauels, A., Lorbacher, K., Meinshausen, N., Etheridge, D. M., Fraser, P. J., Montzka, S. A., Rayner, P. J., Trudinger, C. M., Krummel, P. B., Beyerle, U., Canadell, J. G., Daniel, J. S., Enting, I. G., Law, R. M., Lunder, C. R., O'Doherty, S., Prinn, R. G., Reimann, S., Rubino, M., Velders, G. J. M., Vollmer, M. K., Wang, R. H. J., and Weiss, R.: Historical greenhouse gas concentrations for climate modelling (CMIP6), Geosci. Model Dev., 10, 2057-2116, 2017.
A simple panel on coffee imports for importing countries.
coffee_importscoffee_imports
A data frame with 4530 observations on the following 4 variables.
countrya character vector for the country
membera numeric vector indicating whether the importer is or is not a member of the International Coffee Organization
yeara numeric vector for the year
valuea numeric vector for the coffee imports for all select importing countries (in thousand 60-kg bags)
Data come from the International Coffee Organization, of which I feel I should be a member.
Observations for the People's Republic of China are removed because those can be obtained by adding together the values for "Macao", "Hong Kong", and "China (Mainland)".
The user may want to be mindful about when 0s in the value data are actually communicating that the entry did not exist at the time, or no longer exists. For example, there is no independent Armenia in 1990 (and whatever imports Armenia had are lumped into the USSR value for 1990). Likewise, the 0s for the USSR in 1992 are communicating the USSR no longer exists that year and you should instead look into one of the constituent republics for the information you want. You may want to benchmark this information to some kind of state system membership data.
This is primary commodity price data for coffee (Arabica, Robustas) from 1980 to the present. I manually update these data since FRED's coverage since 2017 has been spotty.
coffee_pricecoffee_price
A data frame with the following 3 variables.
datethe date (year-month)
arabicathe price (monthly average) of mild Arabica, via International Coffee Organization data, in nominal US cents per pound
robustasthe price (monthly average) of Robustas, via International Coffee Organization data, in nominal US cents per pound
Data come from International Monetary Fund (Primary Commodity Prices) and International Coffee Organization. The IMF adds these prices are global and the New York cash price, ex-dock
A data set on select, monthly commodity prices made available by the World Bank in its so-called "pink sheet." These data are potentially useful for applications on data gathering, inflation adjustments, indexing, cointegration, general economic riff-raff, and more.
commodity_pricescommodity_prices
A data frame with the following 11 variables.
datea date
oil_brentcrude oil, UK Brent 38 API ($/bbl)
oil_dubaicrude oil, Dubai Fateh 32 API for years 1985-present; 1960-84 refer to Saudi Arabian Light, 34 API ($/bbl).
coffee_arabicacoffee (ICO), International Coffee Organization indicator price, other mild Arabicas, average New York and Bremen/Hamburg markets, ex-dock ($/kg)
coffee_robustascoffee (ICO), International Coffee Organization indicator price, Robustas, average New York and Le Havre/Marseilles markets, ex-dock ($/kg)
tea_columbotea (Colombo auctions), Sri Lankan origin, all tea, arithmetic average of weekly quotes ($/kg).
tea_kolkatatea (Kolkata auctions), leaf, include excise duty, arithmetic average of weekly quotes ($/kg).
tea_mombasatea (Mombasa/Nairobi auctions), African origin, all tea, arithmetic average of weekly quotes ($/kg).
sugar_eusugar (EU), European Union negotiated import price for raw unpackaged sugar from African, Caribbean and Pacific (ACP) under Lome Conventions, c.I.f. European ports ($/kg)
sugar_ussugar (United States), nearby futures contract, c.i.f. ($/kg)
sugar_worldsugar (World), International Sugar Agreement (ISA) daily price, raw, f.o.b. and stowed at greater Caribbean ports ($/kg).
All data are in nominal USD. Adjust (to taste) accordingly.
Data compiled by the World Bank for its historical data on commodity prices. The oil price data come from a combination of sources, supposedly Bloomberg, Energy Intelligence Group (EIG), Organization of Petroleum Exporting Countries (OPEC), and the World Bank. Data on coffee prices come from Bloomberg, Complete Coffee Coverage, the International Coffee Organization, Thomson Reuters Datastream, and the World Bank. Data on tea prices for Colombo auctions come the from International Tea Committee, Tea Broker's Association of London, Tea Exporters Association Sri Lanka, and the World Bank. Data on tea prices for Kolkata auctions come from the International Tea Committee, Tea Board India, Tea Broker's Association of London, and the World Bank. Tea prices for Mombasa/Nairobi auctions come from African Tea Brokers Limited, International Tea Committee, Tea Broker's Association of London, and the World Bank. EU sugar price data come from International Monetary Fund, World Bank. Sugar price data for the United States come from Bloomberg and World Bank. Global sugar price data come from Bloomberg, International Sugar Organization, Thomson Reuters Datastream, and the World Bank.
This data set effectively deprecates the sugar_price and coffee_price
data sets in this package. Both may be removed at a later point.
A data set of country ISO codes, for my ease and for the ease of my students.
country_isocodescountry_isocodes
A data frame with 249 observations on the following 4 variables.
iso2ca two-character ISO code
iso3ca three-character ISO code
iso3na three-digit numeric ISO code
namean English country name
This is a simple, abbreviated port and rename of the
ISO3_166_1 data in the ISOcodes package.
This is a simple data set provided by Chatterjee and Price (1977, p. 108) that serves as a known example of heteroscedasticity.
CP77CP77
A data frame with 50 observations on the following 6 variables.
statea character vector for the state
regiona character vector for the Census region
urbanpopa numeric vector for the number of residents (per thousand) living in urban areas in 1970
incpca numeric vector for income per capita in 1973
popa numeric vector for residents (per thousand) under 18 years of age in 1974
edexppca numeric vector for per capita public school expenditures in a state, projected for 1975.
I copied these data from the robustbase package.
I just didn't want to make my students install it.
Note: I'm pretty sure "NB" was suppose to be "NE" and that
"DY" is supposed to be "KY". I made those changes.
P. J. Rousseeuw and A. M. Leroy (1987) Robust Regression and Outlier Detection; Wiley, p.110, table 16.
A reduced form of data set for reproducing an analysis on the determinants of Arab public opinion in seven countries toward 13 different countries.
DAPODAPO
A data frame with 91 observations on the following variables.
subjnamea three-character ISO code for the Arab (subject) country
objnamean ALL-CAPS English name for the target/object country
affectan affect rating by the subject country to the object country
capsubthe composite index of national capabilities (capability ratio) of the subject country
capobjthe composite index of national capabilities (capability ratio) of the object country
securtiea dummy variable indicating at least an informal security tie between the subject and object
exportthe volume of exports from the subject to the object
importthe volume of imports to the subject from the object
subgdpthe gross domestic product (GDP) of the subject
islama dummy variable that equals 1 if the object is a predominantly Muslim country
westa dummy variable that equals 1 if the object ia a Western country
Exact coding issues/peculiarities are best addressed by reading the reference article. To maximally reproduce the article's analyses, the user will need to create some variables. The information is here, but you'll need to create a variable for dyadic trade (and as a percentage of the subject's GDP), GDP-adjusted imports, a means to filter out Israel from the analysis, and some of the information reported in Table 1. However, I think this is a learning experience for students.
Furia, Peter A. and Russell E. Lucas. 2006. "Determinants of Arab Public Opinion" International Studies Quarterly 50: 585-605.
An illustrative exercise in never trusting the summary statistics without also visualizing them.
DatasaurusDatasaurus
A data frame with 1,846 observations on the following 3 variables.
datasetthe particular data set, one of 12
xa random variable
yanother random variable
Data were created by Alberto Cairo to illustrate you should always
visualize your data beyond the summary statistics. These are 12 data sets,
in long form, each with a mean of x about 54.26, a mean of y
about 47.83. The standard deviation for x is about 16.76 and the
standard deviation of y is about 26.93. x and y will
correlate weakly, about -.06.
Alberto Cairo, Justin Matejka, George Fitzmaurice
Matejka, Justin and George Fitzmaurice. 2017. “Same Stats, Different Graphs: Generating Datasets with Varied Appearance and Identical Statistics through Simulated Annealing.” ACM SIGCHI Conference on Human Factors in Computing Systems.
A data set on domestic conflict events in 2012 as recorded by the Cross-National Time Series Database. Data exist for teaching about count models.
DCE12DCE12
A data frame with 198 observations on the following 19 variables.
iso2ca two-character ISO code
countrya character name for the country corresponding with the ISO code
assassinationsthe count of assassinations in 2012
strikesthe count of general strikes in 2012
guerwarthe count of guerilla warfare events in 2012
govtcrisesthe count of government crises in 2012
purgesthe count of purges in 2012
riotsthe count of riots in 2012
revolutionsthe count of revolutions in 2012
agdthe count of anti-government demonstrations in 2012
wcithe weighted conflict index in 2012
areathe land area in square kilometers
adultpopthe adult (15+) population (in 1000s)
youthpopthe youth (15-29) population (in 1000s)
gdppcGDP per capita (in constant 2015 USD)
urbanshareurban population over total population (as percentage)
tpoptotal population (in 1000s)
polyarchyelectoral democracy index, an estimate of democracy
perctserpercentage of tertiary school-aged population enrolled in tertiary school
Conflict events data come from the Cross-National Time Series Database. I've used these data before for published papers, but the relative opacity of a data set for yearly purchase comes with a bit of a caveat emptor for the important question of real-world inference.
Data on the democracy estimate and tertiary school enrollment rate come from the Varieties of Democracy project. Democracy estimate for Palestine comes as a simple average of the two Palestinian territories collected by the Varieties of Democracy project. These are West Bank and Gaza. The tertiary school enrollment variable, which originally comes from a data project by Barro and Lee (2013), is "filled" to the referent year from the most recent year available in the data. That would be 2010. It's fine for this purpose.
Population estimates come from the UN Population Division. GDP per capita comes from the World Bank. The estimate of land area (in square kilometers) comes from the CNTS. Country name comes from CNTS as well.
In all but the case of the data from CNTS, and the "filled" case of the tertiary school enrollment variable, the referent year for the data is 2011. Not that anyone is going to care too much for a simple data set like this, but this would be the ol' endogeneity concern.
This should be a data set for a (partial?) replication of Dee's (2004) article on the purported civics returns to education. I use these data for in-class illustration about instrumental variable analyses.
Dee04Dee04
A data frame with 9227 observations on the following 8 variables.
schoolida numeric vector that should be understood as categorical
hispanica numeric vector for if the person is Hispanic
collegea numeric vector for if the person went to college
blacka numeric vector for if the person is black
otherracea numeric vector for if the person is another race
femalea numeric vector for if the person is a woman
registera numeric vector for if the person is registered to vote
distancea numeric vector for the distance to college
I should note I acquired this data set in Mexico City sitting on a two-week program at IPSA-FLACSO Mexico Summer School in 2019. The sample size here (9,227) is about two thousand short of what Dee reports in his article. It'll do, though.
Dee, Thomas S. 2004. "Are there civics returns to education?" Journal of Public Economics 88: 1697–1720
This data set contains the value of the Dow Jones Industrial Average on daily close for all available dates (to the best of my knowledge) from 1885 to the most recent update I feel like including. Extensions shouldn't be too difficult with existing packages.
DJIADJIA
A data frame with the following 2 variables.
datethe date
valuethe value of the the Dow Jones Industrial Average at daily close
Observations before October 7, 1896 are from the single Dow Jones Average.
Observations from October 7, 1896 to July 30, 1914 are from the first DJIA.
Observations before the 1914 closure of the first DJIA in July 1914 come from MeasuringWorth.
Observations from its reopening in Dec. 12, 1914 to Dec. 31 1991 come from
Pinnacle Systems. Observations from Jan. 1, 1992 to the most recent observation come
from a quantmod call.
Samuel H. Williamson, 'Daily Closing Value of the Dow Jones Average, 1885 to Present,' MeasuringWorth, 2019.
Jeffrey A. Ryan and Joshua M. Ulrich, 'quantmod: Quantitative Financial Modelling Framework,' 2018.
These are fatalities (and, in the case of terrorism, casualties as well) for drunk-driving, suicide, and acts of terrorism in the U.S. spanning 1970 to 2018. Only one of these is sufficiently important to command public attention despite being the least severe public bad. Do you want to guess which one?
DSTDST
A data frame with 49 observations on the following 5 variables.
yearthe year
nkilla numeric vector for the number killed in acts of terrorism
terrtotala numeric vector for the number killed or wounded in acts of terrorism
suicidesa numeric vector for the number of suicides
ddfata numeric vector for the number of drunk-driving fatalities
Following my own work in Political Research Quarterly, terror incidents with unknown fatalities or number wounded were imputed to be 1. In those cases, the GTD has reason to believe at least one person died or was wounded, but doesn't know how many. GTD is weird about 1993, so perhaps treat those observations with some care (though it does well to capture the WTC bombing that year). Suicides include only those who passed, not those who survived a suicide attempt. Drunk-driving fatalities seem to include those who were killed in a drunk-driving accident despite not being drunk themselves.
Global Terrorism Database (Sept. 2019 update), Centers for Disease Control, U.S. Department of Transportation
A data set on the apparent economic benefits of post-conflict justice
EBJEBJ
A data frame with 95 observations on the following 12 variables.
testnewid_lagan apparent identifier variable, of some description
ccodea Correlates of War(?) state code for the location of a conflict
idan apparent identifier variable, of some description
pcja dummy variable for whether there was some kind of post-conflict justice institution created after a conflict
fdia variable on net FDI inflows over a 10-year period after a conflict (in millions USD)
econ_sizeGDP, as an estimate of economic size
econ_develGDP per capita, as an estimate of economic development
econ_growthGDP per capita change, as an estimate of economic growth
kaopenKAOPEN index score, as an estimate of capital openness
xrexchange rate fluctuations, as an indicator of exchange rate instability
lflabor force size
lifeexpaverage life expectancy for women, in years
Data are taken Appell and Loyle's (2012) replication data set. Users should read their article in Journal of Peace Research for more information about the topic, the stake, and how the data were collected. This is just a simple, reduced form of the data they make available that is minimally sufficient for reproducing the first model of their Table I.
Appell, Benjamin J. and Cyanne E. Loyle. 2012. "The Economic Benefits of Justice: Post-conflict Justice and Foreign Direct Investment" 49(5): 685–99.
You've all seen these before. These are the "eight schools" that everyone gets when being introduced to Bayesian programming. Here are the full data for your consideration, which you can use instead of awkwardly searching where the data are and copy-pasting them as a list. Every damn time, Steve.
eight_schoolseight_schools
A data frame with 8 observations on the following 6 variables.
schoola letter denoting the school
num_treatthe number of students in the school receiving the treatment
num_controlthe number of students in the school in the control group
estthe estimated treatment effect
sethe standard error of the effect estimate
rvarthe residual variance
Data copy-pasted from Table 1 in Rubin (1981).
Rubin, Donald B. 1981. "Estimation in Parallel Randomized Experiments." Journal of Educational Statistics 6(4): 377-401.
A simple data set on education and state-level (+ DC) turnout in the 2016 presidential election. This is inspired by what Pollock (2012) does in his book.
election_turnoutelection_turnout
A data frame with 51 observations on the following 13 variables.
yearthe year of the presidential election (2016)
statethe state abbreviation
regionthe state's Census region
divisionthe state's Census division
turnouthovoter turnout for the highest office as percent of voting-eligible population (VEP)
perhsedthe percentage of the state that completed high school
percoledthe percentage of the state that completed college
gdppercapan estimate of the state's GDP per capita
ssis it a “swing state?”
trumpwdid Trump win the state?
trumpsharethe share of the vote Trump received
sunemprthe state-level unemployment rate entering Nov. 2016
sunempr12mdthe state-level unemployment rate (12-month difference) entering Nov. 2016. Higher values indicate the unemployment rate is increasing entering Nov. 2016 relative to what it was entering Nov. 2015.
gdpan estimate of the state's GDP
Data were created in early 2017 for an upper-division course on quantitative methods. Educational attainment and division/region data come from the Census. Voter turnout/share data come from the Elections Project at George Mason University. GDP per capita estimates come from Bureau of Economic Analysis. Unemployment data come from the Bureau of Labor Statistics and code to generate it was derived from a forthcoming publication of mine.
A data set on the odds for relegation and winning the table for English Premier League clubs for the 2024-25 season. Data are useful for illustrating what exactly odds are.
epl_oddsepl_odds
A data frame with 20 observations on the following 7 variables.
cluba character communicating the name of the club in the Premier League
bet365_ra numeric vector for the odds against relegation, by way of Bet365
betfair_ra numeric vector for the odds against relegation, by way of Betfair
unibet_ra numeric vector for the odds against relegation, by way of Unibet
bet365_wa numeric vector for the odds against winning the table, by way of Bet365
betfair_wa numeric vector for the odds against winning the table, by way of Betfair
unibet_wa numeric vector for the odds against winning the table, by way of Unibet
Data come oddschecker.com as of Oct. 20, 2024, assuming these
are preseason odds. Raw data are available on the project's website for your
consideration. Don't bet on sports, unless you've been visited by Biff Tannen
from the future.
Fractions have been converted into decimals for ease of maintaining the data. Raw odds (in fraction form) for those clubs most likely to be relegated are available in the raw data files on the project's Github.
Odds are typically(?) communicated as "odds against" in the sports betting world. It's why the highest odds for relegation and lowest odds for winning coincide with the biggest, most successful clubs. Context clues help, and are useful for understanding what these odds are saying.
It's possible that the language "win outright" is doing some heavy-lifting in how Bet365 lets you place wagers on winning the table.
I'm also aware of the reason these odds do not sum to 1, and in fact exceed
If anything, I think "overrounding" is its own pedagogical tool for why odds can be wonky things to learn in relation to its use in the statistical modeling context.
Data from the International Monetary Fund for the export quality and unit/trade value of passenger cars for all available countries and years from 1963 to 2014.
eq_passengercarseq_passengercars
A data frame with 60424 observations on the following 6 variables.
countrya character vector for the country/area.
ccodea numeric vector for the Correlates of War country code.
categorya factor with levels Export Quality Index,
Export quality 95 percent interval - lower bound,
Export quality 95 percent interval - upper bound
Unit value of exports, Unit value 95 percent interval - lower bound,
Unit value 95 percent interval - upper bound,
Trade value of exports
typea factor with levels 51. Transport equipment, Passenger cars.
This is a constant. I just felt like making it a factor.
yeara numeric vector for the year
valuea numeric vector for the value of the particular category.
This is a simple data set to illustrate the use of sampling weights from the European Social Survey.
ESS10NOESS10NO
A data frame with 1,411 observations on the following 24 variables.
cntrya character vector with Norway's two-character ISO code
idnoa numeric identifier for the individual respondent
regiona character for one of six regions recorded by the European Social Survey
inwdsa date-time vector for the start of the interview
inwdea date-time vector for the end of the interview
dweighta design weight
pspwghta post-stratification weight, including the design weight
pweighta population size weight
anweightan analysis weight
probthe sampling probability
stratumthe sampling stratum
psuthe primary sampling unit
eu_votea character vector indicating how a respondent would vote if given a vote on joining the European Union
brnnorgea dummy variable indicating whether respondent was born in Norway or not
ageaa numeric vector for the respondent's age in years
imbgecoa numeric vector for if respondent thinks immigrants are generally good or bad for Norway's economy. Higher values = good
imueclta numeric vector for if respondent thinks immigrants enrich or undermine Norway's culture. Higher values = enrich more than undermine
imwbcnta numeric vector for if respondent thinks immigrants make Norway a better place to live. Higher values = better place to live
femalea numeric vector for whether the respondent is a woman
eduyrsa numeric vector for total years of education for the respondent
uemplaa numeric vector for whether the respondent is currently unemployed but seeking work
polinta dummy variable indicating political interest. 1 = very or quite interested. 0 = hardly or not at all interested.
hinctntaa numeric vector for household income in deciles
lrscalea numeric vector for the ideology of the respondent on an 11-point scale, from 0 to 10
You'll want to convert the eu_vote variable into something usable.
Possible values include "Remain Outside", "Join EU", "Don't Know", "Not Eligible",
"Blank Ballot", "Refuse to Answer", "Wouldn't Vote". Perhaps it's reasonable
to make this a dummy variable comparing those who want to join versus those
who want Norway to remain outside the European Union.
The data are edition 2.2 of the 10th round of European Social Survey, which was released for public consumption on 21 December 2022.
European Social Survey, Round 10
This is a replication data originally set to accompany a blog post and presentation to students at the University of Nottingham in March 2020. However, COVID-19 led to the cancellation of the talk.
ESS9GBESS9GB
A data frame with 1,905 observations on the following 19 variables.
namea character for the name of the survey
essrounda numeric for the ESS round
editiona character for the particular edition of the ESS round
idnoa numeric/unique identifier
cntrya character vector for the country (i.e. the UK)
regiona character vector for the region of the UK the respondent lives
brncntra numeric vector for if the respondent was born in the UK
stintrvwa Date for the interview start date
endintrvwa Date for the interview end date
imbgecoa numeric vector for if respondent thinks immigrants are generally good or bad for UK's economy. Higher values = good
imueclta numeric vector for if respondent thinks immigrants enrich or undermine UK's culture. Higher values = enrich more than undermine
imwbcnta numeric vector for if respondent thinks immigrants make UK a better place to live. Higher values = better place to live
immigsenta numeric vector for immigration sentiment (i.e. imbgeco +
imueclt + imwbcnt). Higher values = more pro-immigration sentiment
ageaa numeric vector for the respondent's age in years
femalea numeric vector for whether the respondent is a woman
eduyrsa numeric vector for total years of education for the respondent
uemplaa numeric vector for whether the respondent is currently unemployed but seeking work
hinctntaa numeric vector for household income in deciles
lrscalea numeric vector for the ideology of the respondent on an 11-point scale, from 0 to 10
See accompanying blog post at http://svmiller.com/blog/2020/03/what-explains-british-attitudes-toward-immigration-a-pedagogical-example/.
European Social Survey, Round 9
This is a sample data set cobbled from the fifth round of European Social Survey data for Belgium. It offers a means to do a basic replication of some of Chapter 5 of The SAGE Handbook of Regression Analysis and Causal Inference.
ESSBE5ESSBE5
A data frame with 1704 observations on the following 10 variables.
essrounda numeric for the ESS round
editiona character for the edition number of the fifth round
idnoa numeric id number
cntrya character vector for the country (i.e. Belgium, or BE)
trstplca numeric vector for trust in the police on an 11-point scale. Higher values indicate more trust. 0 = "no trust at all". 10 = "complete trust"
ageaa numeric vector for the respondent's age
femalea numeric vector for whether the respondent is a woman or not.
eduyrsa numeric vector for years of education.
hincfela numeric vector for the respondent's feeling about their household income. 1 = "living comfortably", 2 = "coping on present income", 3 = "difficult on present income", 4 = "very difficult on present income"
plcpvcra numeric vector for how successful police are at preventing crimes in a country on an 11-point scale. 0 = "extremely unsuccessful". 10 = "extremely successful."
See Chapter 5 of The SAGE Handbook of Regression Analysis and Causal Inference for more information.
European Social Survey (Round 5)
A data set on an April 2024 roll call vote to extend an emergency free trade agreement with Ukraine.
eu_ua_fta24eu_ua_fta24
A data frame with 705 observations on the following 9 variables.
member_ida numeric identifier for a member of the European Parliament
first_namea first name of the member of the European Parliament
last_namea last name of the member of the European Parliament
positiona character vector indicating the member's position on the measure ("For", "Against", "Did Not Vote", or "Abstain")
iso2ca two-character ISO code for the country the member represents
countryan English country name for the country the member represents
group_codean acronym/code for the coalition of the member
group_labela character vector for the full name of the coalition of the member
group_short_labela short label for the coalition of the member
Information about the exact measures are available from the European
Union. Search: A9-0077/2024. Data in question are the raw data
made available by HowTheyVote.eu
A data set taken from Eurostat's glossary on codes and country classifications.
eurostat_codeseurostat_codes
A data frame on the following 3 variables.
countryan English country/territorial unit name
iso2ca two-character code for the country/territorial unit
cata category indicator for the country/territorial unit. See Details section for more.
The ISO two-character code for Kosovo is not "XK". XK is a "user assigned" ISO 3166 code that is not used by the International Organization for Standardization, but is nevertheless in wide use by entities like the European Commission. To the best of my knowledge, Kosovo's official ISO classification is still what it was when it was a subdivision of Serbia/Yugoslavia.
A glossary on Eurostat provides the following category entries included in this data frame. "EU" is an European Union member. "EFTA" are countries outside the European Union, but still included in the free trade agreement. "UK" is the United Kingdom, because they left. "EUCC" is a category for European Union candidate countries. "PC" are potential candidates. European Union expansion led to the delineation of neighboring states to "South" and "East" as part of the European Neighbourhood Policy (ENP). "OEC" stands for "Other European Countries", but is effectively a simple indicator for Russia.
European Union membership by accession date
eustateseustates
A data frame with 28 observations on the following 3 variables.
datea date indicating accession
countrya character vector for the country
iso2ca character vector for iso2c
Data come from the European Union's website.
This is a hypothetical universe of schools in a given territorial unit,
patterned off the apipop data available in the survey package.
fakeAPIfakeAPI
A data frame with 10000 observations on the following 8 variables.
uida numeric vector as a unique identifier for schools
schooltypea character vector for school type. E = elementary school. M = middle school. H = high school
countya character vector for the county, named after an Ohio State All-American. “County” incidence is weighted by how many All-American honors the Ohio State player had. It's my fake data. You make your own if you have a problem with it.
communitya character vector for the school's community, either rural, suburban, or urban.
apia numeric vector vector an academic performance index for the school
mealsa numeric vector for the percentage of school students eligible for subsidized meals
colgrada numeric vector for the percentage of school parents with college degrees
fullquala numeric vector for the percentage of the school with teachers that are fully qualified
sbasea numeric vector for some base differences between schools, patterned off the school type means for api00 in the apipop data.
cbasea numeric vector for some base differences between counties, randomly drawn from a uniform distribution
ea numeric vector for random errors
These data were generated for a blog post on my website.
Miller, Steven V. 2020. "Some Parlor Tricks with Survey-Type Analyses in R." URL: http://svmiller.com/blog/2020/08/some-parlor-tricks-with-survey-type-analyses-in-r/
This is a toy ("fake") data set I might use to illustrate the so-called curvilinear effect of age on happiness.
fakeHappinessfakeHappiness
A data frame with 1000 observations on the following 8 variables.
agea numeric vector for age.
femalea numeric that equals 1 if the respondent is a woman
collegeeda numeric vector that equals 1 if the respondent says s/he has a college degree
famincra numeric vector for the respondent's household income. Ranges from 1 to 12.
bornagaina numeric vector for whether the respondent self-identifies as a born-again Christian.
erandom noise, generated from a normal distribution with a mean of 0 and a standard deviation of 3
happyan arbitrary happiness variable. See details for its construction
z_happythe same arbitrary happiness variable, scaled to have a mean of 0 and a standard deviation of 1. This makes it seem more "latent".
Data are randomly sampled from the TV16 data set in the same package
for the age, female, college education, family income, and born-again variables.
Thereafter, I created an arbitrary "happiness" variable that is equal to
100 - .95*age + .01*(age^2) + .25*female + .05*famincr + .1*bornagain + e. The
data are not supposed to be realistic, per se. They're supposed to be functional for this
purpose.
This is a simple fake data set to illustrate a logistic regression.
fakeLogitfakeLogit
A data frame with 10000 observations on the following 2 variables.
xa five-item functionally ordered categorical variable
ya binary variable that is either 0 or 1
The data are generated such that the outcome y is a logistic
function of the x variable and come from a rbinom() call. The
estimated natural logged odds of y when x is 0 is -2.8. Each
unit increase in x is simulated to increase the natural logged odds of
y by 1.4. This example is very much patterned off a similar fake data
set that Pollock (2012) uses to teach about logistic regression. In his case,
x is a stand-in for hypothetical education categories and y is
whether this fake person voted or not.
This is a toy (i.e. "fake") data set created by the fabricatr package.
There are 100 observations for 25 hypothetical countries. The outcome y
is a linear function of a baseline for each hypothetical country, plus a yearly
growth trend as well as varying growth errors for each country. x1 is
supposed to have a linear effect of .5 on y, all things considered.
x2 is supposed to have a linear effect of 1 on y for each unit
change in x2, all things considered.
fakeTSCSfakeTSCS
A data frame with 2500 observations on the following 8 variables.
yeara numeric vector for the year
countrya character vector for the country
ya numeric vector for the outcome.
x1a continuous variable
x2a binary variable
basea numeric vector for the baseline starting point for each country
growth_unitsa numeric vector for the growth units for each country
growth_errora numeric vector for the growth errors for each country
x1 is generated by a normal distribution with a mean of 5 and a standard
deviation of 2. x2 is drawn from a Bernoulli distribution with a
probability of .5 of observing a 1.
This is a toy (i.e. "fake") data set created by the fabricatr package.
There are 100 observations. The outcome y is a linear function
of 20 + (.25 * year) + .(25 * x1) + (1 * x2) + e. This clearly implies
some autocorrelation in the data. I.e. it's a time-series.
fakeTSDfakeTSD
A data frame with 100 observations on the following 5 variables.
yearthe year
yan outcome
x1a continuous variable
x2a binary variable
erandomly generated errors
Errors are random-normal with a mean of 0 and a standard deviation of 1.
x1 is generated by a normal distribution with a mean of 5 and a standard
deviation of 2. x2 is drawn from a Bernoulli distribution with a
probability of .5 of observing a 1.
This is the yearly rate of gun homicides per 100,000 people in the population, selecting on "Western" countries of interest.
ghp100kghp100k
A data frame with 561 observations on the following 3 variables.
countrythe country
yearthe year
valuea numeric vector for the estimated rate of gun homicide per 100,000 people
The reported, or calculated annual crude rate of completed, intentional homicide committed with a firearm, per 100,000 population, in years descending.
Where a jurisdiction's published count of 'annual homicide' includes cases of attempted (uncompleted) homicide, these figures have been disaggregated wherever possible.
In the United States, this category is confused by inaccurate and conflicting data published, suppressed or labeled as unreliable by the Centers for Disease Control and Prevention (CDC) and the Federal Bureau of Investigation (FBI). Suppression can result in zero values where in fact homicides did occur.
Incomplete classification by local agencies can also result in a significant proportion of events being categorized as 'unknown cause' or similar.
Before quoting these datasets, please follow the citation links for a description of the considerable differences between them and the reasons for data suppression.
Where a rate is calculated by GunPolicy.org, a matched population
estimate is also cited.
The aforementioned details come, copied and pasted, from GunPolicy.org.
As of my most recent check of these data (April 2024), this agency appeared
to close due to lack of funding. This is unfortunate, but it is worth noting
for matters of reproducibility and the use of these data in applied research
questions.
This is a data set for replicating Ghobarah et al. (2004), a reduced form of what they make available on Dataverse for replication. Variables have been renamed for legibility.
GHR04GHR04
A data frame with 182 observations on the following 15 variables.
countrya character vector denoting a country name
iso3ca three-character ISO code for the country
pubhlthexppgdpa numeric vector for public health expenditures as a percentage of GDP
totexphltha numeric vector for total expenditures on health
halea numeric vector for health adjusted life expectancy (in years)
log_gdppca numeric vector for (log-transformed) GDP per capita
ginia numeric vector for income inequality
log_educa numeric vector for (log-transformed) educational attainment
log_vanhanena numeric vector for (log-transformed) racial-linguistic-religious heterogeneity
rivalrya dummy variable indicating the presence of an enduring international rivalry for the country
politya numeric vector communicating a Polity score, as a measure of the democratic nature of the country's regime
prvhlthexpgdpa numeric vector for private spending on health as a percentage of GDP
urban_growtha numeric vector for the pace of urbanization
cwdeathsa numeric vector for civil war deaths
contig_cwa dummy variable communicating whether there is a civil war in a geographically contiguous territory
The three-character ISO code is the only new addition to the data. I add this because the country names they have in the data are not neat and may lead users astray if they wanted to search for a specific observation. The ISO code for Yugoslavia (Serbia and Montenegro) around this time was "SCG".
The data the authors make available come with no .do file to indicate what
exactly they used. Some forensic work based on the descriptive statistics they
mention led to this reduced form of their data, which almost perfectly replicates
their results. The differences are typically in the hundredths, and often in
the thousandths, and should be considered "good enough" for replication
purposes. The only real confusion on my end is why I ended up with one more
observation than they report in their analyses. This suggests one (or more?)
of their variables they use has an NA, but I have no way of knowing what it
could be.
Ghobarah, Hazem Adam, Paul Huth, and Bruce Russett. 2004. "Comparative Public Health: The Political Economy of Human Misery and Well-Being" International Studies Quarterly 48: 73-94
This is a toy data set derived from the General Social Survey that I intend to use for several purposes. First, the battery of abortion items can serve as toy data to illustrate mixed effects modeling as equivalent to a one-parameter (Rasch) model. Second, I include some covariates to also do some basic regressions. I think abortion opinions are useful learning tools for statistical inference for college students. Third, there's a time-series component as well for understanding how abortion attitudes have changed over time.
gss_abortiongss_abortion
A data frame with 64,814 observations on the following 18 variables.
ida unique respondent identifier
yearthe survey year
agethe respondent's age in years
racethe respondent's race, as character variable
sexthe respondent's gender, as character variable
hispaniccatthe respondent's Hispanic ethnicity, as character variable
educhow many years the respondent spent in school
partyidthe respondent's party identification, as character variable
relactivthe self-reported religious activity of the respondent on a 1:11 scale
abanya binary variable that equals 1 if the respondent thinks abortion should be legal for any reason. 0 indicates no support for abortion for any reason.
abdefecta numeric vector that equals 1 if the respondent thinks abortion should be legal if there is a serious defect in the fetus. 0 indicates no support for abortion in this circumstance.
abnomorea numeric vector that equals 1 if the respondent thinks abortion should be legal if a woman is pregnant but wants no more children. 0 indicates no support for abortion in this circumstance.
abhltha numeric vector that equals 1 if the respondent thinks abortion should be legal if a pregnant woman's health is in danger. 0 indicates no support for abortion in this circumstance.
abpoora numeric vector that equals 1 if the respondent thinks abortion should be legal if a pregnant woman is poor and cannot afford more children. 0 indicates no support for abortion in this circumstance.
abrapea numeric vector that equals 1 if the respondent thinks abortion should be legal if the woman became pregnant because of a rape. 0 indicates no support for abortion in this circumstance.
absinglea numeric vector that equals 1 if the respondent thinks abortion should be legal if a pregnant woman is single and does not want to marry the man who impregnated her. 0 indicates no support for abortion in this circumstance.
pidpartyid recoded so that 7 = NA
hispanica dummy variable that equals 1 if the respondent is any way Hispanic
Data include all General Social Survey observations from 1972 to 2018 for these variables. Be mindful of missing data.
This is a toy data set that collects attitudes on toward national spending for various things in the General Social Survey for 2018. I use these data for in-class illustration about ordinal variables and ordinal models.
gss_spendinggss_spending
A data frame with 2348 observations on the following 33 variables.
yeara numeric constant for the GSS survey year (2018)
ida unique identifier for the survey respondent
agea numeric vector for the age of the respondent (min: 18, max: 89)
sexa numeric vector for the respondent's sex (1 = female, 0 = male)
educa numeric vector for the highest year of school completed (min: 0, max: 20)
degreea numeric vector for the respondent's highest degree (0 = did not graduate high school, 1 = high school, 2 = junior college, 3 = bachelor degree, 4 = graduate degree)
racea numeric vector for the respondent's race (1 = white, 2 = black, 3 = other)
rincom16a numeric vector for the respondent's yearly income (min: 1 (under $1,000), max: 26 ($170,000 or over))
partyida numeric vector for the respondent's party identification on the familiar seven-point scale. NOTE: D to R partisanship in this variable goes from 0 to 6. 7 = supporters of other parties. You may want to recode this if you want an interval-level measure of partisanship.
polviewsa numeric vector for the respondent's ideology (min: 1 (extremely liberal), max: 7 (extremely conservative))
xnorcsiza numeric vector for the NORC size code. This is a measure of what kind of area in which the respondent took the survey (i.e. lives). 1 = city, greater than 250k residents. 2 = city, between 50k-250k residents. 3 = suburbs of a large city. 4 = suburbs of a medium-sized city. 5 = unincorporated area of a large city. 6 = unincorporated area of a medium city. 7 = city, between 10-50k residents. 8 = town, greater than 2,500 residents. 9 = smaller areas. 10 = open country.
newsa numeric vector for how often the respondent reads the newspapers. 1 = everyday. 2 = a few times a week. 3 = once a week. 4 = less than once a week. 5 = never.
wrkstata numeric vector for the respondent's work status. 1 = working full-time. 2 = working part-time. 3 = temporarily not working. 4 = unemployed/laid off. 5 = retired. 6 = in school. 7 = house-keeping work. 8 = other.
natspaca numeric vector for attitudes toward spending on the space program. See details below for this variable and all other variables beginning with nat.
natenvira numeric vector for attitudes toward spending on improving/protecting the environment.
natheala numeric vector for attitudes toward spending on improving/protecting the nation's health.
natcitya numeric vector for attitudes toward spending on solving the big city's problems.
natcrimea numeric vector for attitudes toward spending on halting the "rising crime rate." This question is subtly hilarious.
natdruga numeric vector for attitudes toward spending on dealing with drug addiction.
nateduca numeric vector for attitudes toward spending on improving the nation's education system.
natracea numeric vector for attitudes toward spending on improving the condition of black people.
natarmsa numeric vector for attitudes toward spending on the military/armaments/defense.
nataida numeric vector for attitudes toward spending on foreign aid.
natfarea numeric vector for attitudes toward spending on welfare.
natroada numeric vector for attitudes toward spending on highways and bridges.
natsoca numeric vector for attitudes toward spending on social security.
natmassa numeric vector for attitudes toward spending on mass transportation.
natparka numeric vector for attitudes toward spending on parks and recreation.
natchlda numeric vector for attitudes toward spending on assistance for child care.
natscia numeric vector for attitudes toward spending on scientific research.
natenrgya numeric vector for attitudes toward spending on alternative sources of energy.
sumnata numeric vector for the sum total of responses to all the aforementioned spending variables (i.e. those that begin with nat). This creates an interval-ish measure with a nice and mostly normal distribution.
sumnatsoca numeric vector for the sum of all responses toward various "social" prompts (i.e. natenvir, natheal, natdrug, nateduc, natrace, natfare, natroad, natmass, natpark, natsoc, natchld). This creates an interval-ish measure with a mostly normal (but small left skew) distribution.
For all the variables beginning with nat, note that I rescaled the original data so that -1 = respondent thinks country is spending too much on this topic, 0 = respondent thinks country is spending "about (the) right" amount, and 1 = respondent thinks country is spending too little on this topic. I do this to facilitate reading each nat prompt as increasing support for more spending (the extent to which increasing values means the respondent thinks the country spends too little on a given prompt). I think this is more intuitive.
Also, the natspac, natenvir, natheal, natcity, natcrime, natdrug, nateduc, natrace, natarms, nataid, and natfare have "alternate" prompts in later GSS waves in which a subset of respondents get a slightly different prompt. For example, one set of respondents for natcity gets a prompt of "Solving the problems of the big cities" (the legacy prompt) whereas another set of respondents gets a prompt of "Assistance to big cities" (typically noted as "version y" in the GSS). I, perhaps problematically if I were interested in publishing analyses on these data, combine both prompts into a single variable. I don't think it's a huge problem for what I want the data to do, but FYI.
General Social Survey, 2018
Wage data from the General Social Survey (1974-2018) to illustrate wage discrepancies by gender (while also considering respondent occupation, age, and education).
gss_wagesgss_wages
A data frame with 11 variables:
yearthe survey year
realrincthe respondent's base income (in constant 1986 USD)
agethe respondent's age in years
occ10respondent's occupation code (2010)
occrecoderecode of the occupation code into one of 11 main categories
prestg10respondent's occupational prestige score (2010)
childsnumber of children (0-8)
wrkstatthe work status of the respondent (full-time, part-time, temporarily not working, unemployed (laid off), retired, school, housekeeper, other)
genderrespondent's gender (male or female)
educcatrespondent's degree level (Less Than High School, High School, Junior College, Bachelor, or Graduate)
maritalcatrespondent's marital status (Married, Widowed, Divorced, Separated, Never Married)
For further details, see the GSS Data Explorer at the National Opinion Research Center (NORC) at the University of Chicago. Consult https://census.gov for more information about occupation codes.
A data set for a canonical case of a Simpson's paradox, useful for in-class instruction on the topic.
Guber99Guber99
A data frame with 50 observations on the following 8 variables.
statea character vector for the state
expendppa numeric vector for the current expenditure per pupil in average daily attendance in public elementary and secondary schools, 1994-95 (in thousands of dollars)
ptratioa numeric vector for the average pupil/teacher ratio in public elementary and secondary schools, Fall 1994
tsalarya numeric vector for the estimated average annual salary of teachers in public elementary and secondary schools, 1994-95 (in thousands of dollars)
perctakersa numeric vector for the percentage of all eligible students taking the SAT, 1994-95
verbala numeric vector for the average verbal SAT score, 1994-95
matha numeric vector for the average math SAT score, 1994-95
totala numeric vector for the average total SAT score, 1994-95
Guber, Deborah Lynne. 1999. "Getting What You Pay For: The Debate Over Equity in Public School Expenditures." Journal of Statistics Education 7(2).
This is perhaps the canonical data set for illustrating the ecological fallacy.
illiteracy30illiteracy30
A data frame with 40 observations on the following 11 variables.
statea character for the state
popa numeric vector for the total population
pop_ila numeric vector for the total population that is illiterate
nwhitea numeric vector for the total native white population
nwhite_ila numeric vector for the total native white population that is illiterate
fpwhitea numeric vector for the total white population with "foreign or mixed parentage"
fpwhite_ila numeric vector for the total white population with "foreign or mixed parentage" that is illiterate
fbwhitea numeric vector for the total foreign-born white population
fbwhite_ila numeric vector for the total foreign-born white population that is illiterate
blacka numeric vector for the total black population.
black_ila numeric vector for the total black population that is illiterate
All population totals reflect those 10 years or older. The 1930 Census (along with Robinson (1950)) uses "negro" in lieu of black, but the variable names here eschew that older label. Note that some states are not yet states in the 1930 Census.
U.S. Census Bureau (1933). Fifteenth Census of the United States: 1930. Population, Volume II.
Grotenhuis, Manfred Te, Rob Eisinga, and SV Subramanian. 2011. "Robinson's Ecological Correlations and the Behavior of Individuals: methodological corrections." Internatoinal Journal of Epidemiology 40(4): 1123-25.
Robinson, WS. 1950. "Ecological Correlations and the Behavior of Individuals." American Sociological Review 15(3): 351–57.
A data set based on summary information provided in Inglehart's (2003) article in PS: Political Science & Politics. These data would be from the article itself and only indirectly from the raw World or European Values Survey.
inglehart03inglehart03
A data frame with 77 observations on the following 4 variables.
state_yearthe state year and survey year, as provided in the article
havedemthe percentage of respondents saying having a democratic political system is "very good" or "good"
strongleaderthe percentage of respondents saying having a strong leader unencumbered by elections or parliaments is "very good" or "good"
muslima dummy variable that equals 1 if Inglehart codes the state as being a "predominently Islamic society"
Data manually entered based on Table 1 and Table 2 in Inglehart's (2003) article.
Inglehart, Ronald. 2003. "How Solid is Mass Support for Democracy—And How Can We Measure It?" PS: Political Science & Politics 36(1): 51–57.
A data set on democracy and economic development for 48 countries that Lipset (1959) first described.
Lipset59Lipset59
A data frame with 48 observations on the following 11 variables.
countrya character country for an English country name
cata category for the country by their region and level of democracy
iso3ca three-character ISO code
wbgdp2011estan estimated gross domestic product in 2011 USD
wbpopestan estimated population size
unpopa population size (in thousands)
uninca national income (in millions)
unincpca national income per capita
xm_qudsesta "Quick UDS" estimate of democracy on a latent scale (see details)
v2x_polyarchythe Varieties of Democracy "polyarchy" estimate (see details)
polity2the polity2 score from the Polity project (see details)
The three variables with the prefix of un nominally come from the
United Nations Statistical Division for 1949/1950, but are actually retrieved
from Andic and Peacock (1961). Andic and Peacock (1961) note you should be
skeptical of Soviet-style calculations of national income and thus don't
include it in the data they make available.
Anything else is explicitly benchmarked to 1950 as a referent year. The GDP and population estimates come by way of Anders et al. (2020). You can manually create your own GDP per capita variable here because the GDP is demarcated in dollars and the population size is in units of 1. Take one and divide it over the other.
The democracy variables are all unique in their own way. The "Quick UDS" estimates are generated to be latent and, globally, have a mean that approximates 0 and a standard deviation that approximates 1. In the regression context, that would mean a coefficient would communicate something like a magnitude change across a standard deviation on the scale. The "polyarchy" estimate has a theoretical minimum of 0 and a theoretical maximum of 1. In the regression context, that would mean a coefficient communicates a min/max effect. The Polity project estimate comes from a usual scale of -10 to 10 and a regression coefficient communicates something much less exotic. It's a unit change on this scale.
In all cases, higher values of democracy = more "democraticness", for lack of a better term.
Anders, Therese, Christopher J. Fariss, and Jonathan N. Markowitz. 2020. "Bread Before Guns or Butter: Introducing Surplus Domestic Product (SDP)" International Studies Quarterly 64(2): 392–405.
Andic, Suphan and Alan T. Peacock. 1961. "The International Distribution of Income, 1949 and 1957." Journal of the Royal Statistical Society. Series A (General) 124(2): 206-218.
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, M. Steven Fish, Adam Glynn, Allen Hicken, Anna Luhrmann, Kyle L. Marquardt, Kelly McMann, Pamela Paxton, Daniel Pemstein, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Agnes Cornell, Lisa Gastaldi, Haakon Gjerlow, Valeriya Mechkova, Johannes von Romer, Aksel Sundtrom, Eitan Tzelgov, Luca Uberti, Yi-ting Wang, Tore Wig, and Daniel Ziblatt. 2020. "V-Dem Codebook v10" Varieties of Democracy (V-Dem) Project.
Lipset, Seymour Martin. 1959. "Some Social Requisites of Democracy: Economic Development and Political Legitimacy" American Political Science Review 53(1): 69-105.
Marshall, Monty G., Ted Robert Gurr, and Keith Jaggers. 2017. "Polity IV Project: Political Regime Characteristics and Transitions, 1800-2017." Center for Systemic Peace.
Marquez, Xavier, "A Quick Method for Extending the Unified Democracy Scores" (March 23, 2016). doi:10.2139/ssrn.2753830
Pemstein, Daniel, Stephen Meserve, and James Melton. 2010. "Democratic Compromise: A Latent Variable Analysis of Ten Measures of Regime Type." Political Analysis 18(4): 426-449.
These data contain monthly mean temperature anomalies expressed as deviations from the corresponding 1951-1980 means. They are useful for showing how we can measure climate change.
LOTILOTI
A data frame with 1,716 observations on the following 2 variables.
datea date, mostly to contain information for the year and month
valuethe mean temperature anomaly as deviation from corresponding 1951-1980 mean
Data are updated through most recent month, at least for last time I updated it.
Data represent combined land-surface air and sea-surface water temperature anomalies.
Of note: the day value in the date column has no real value.
It was just a way of combining data that are aggregated by year and month.
National Aeronautics and Space Administration's Goddard Institute for Space Studies.
These data are a monthly time-series of changes in the consumer price index relative to a Dec. 1997 starting date for televisions, computers, and related items. I use this as in-class illustration that globalization has made consumer electronics cheaper across the board for Americans.
LTPTLTPT
A data frame with 1,704 observations on the following 3 variables.
datea date
categorythe particular category (e.g. all items, televisions, etc.)
valuethe consumer price index (Dec. 1997 = 100)
This is a web-scraping job from the U.S. Bureau of Labor Statistics. Post is titled "Long-term price trends for computers, TVs, and related items" and was published on Oct. 13, 2015.
U.S. Bureau of Labor Statistics.
"Let Them Watch TV": These data contain price indices for various items for the general urban consumer. Categories include medical services, college tuition, college textbooks, child care, housing, food and beverages, all items (i.e. general CPI), new vehicles, apparel, and televisions. The base period in value was originally the 1982-4 average, but I converted the base period to January 2000. I use these data for in-class discussion about how liberalized trade has made consumer electronics (like TVs) fractions of their past prices. Yet, young adults face mounting costs for college, child-raising, and health care that government policy has failed to address.
LTWTLTWT
A data frame with 2377 observations on the following 3 variables.
datea date
categorya factor for the particular category
valuethe price index. Base: January 2000
Inspiration comes from a blog post titled "Chart of the day (century?): Price changes 1997 to 2017", which was published by the American Enterprise Institute on Feb. 2, 2018.
Bureau of Labor Statistics, via the blscrapeR package.
A data set on the various federal minimum wage rates.
min_wagemin_wage
A data frame with 23 observations on the following 5 variables.
datea date for when a new minimum wage was introduced
wagethe (nominal) value of the wage
Data come from the Department of Labor. Wages are taken from wage adjustments from the 1938 act.
Department of Labor
A data set on the correlates of government control in 26 provinces in South Vietnam, to replicate a study by Mitchell (1968).
Mitchell68Mitchell68
A data frame with 26 observations on the following 9 variables.
ida numeric vector (a simple identifier)
provincea character vector for the name of the province
gca numeric vector for government control in the province (as a percent)
oola numeric vector for owner-operated land (as a percent)
cvlhsa numeric vector for the coefficient of variation of the distribution of land-holdings, by size
vla numeric vector for Vietnamese land, subject to transfer (as a percent of all land)
fla numeric vector for French land, subject to transfer (as a percent of all land)
ma numeric vector for area of mobility
pda numeric vector for population density (per square kilometer)
The data are gathered from Table 1 in the document. You should also read the article for more information as to what's happening and for what purpose. Mitchell (1968) is quite clear about where else he's getting these data. Much of what follows can be discerned in the first few pages of Mitchell (1968), which jumps right into a conversation about research design after a brief introduction.
Province names are taken "as is" from Mitchell (1968). Since South Vietnam no longer exists, and these observations are about 60 years old, some of these province names may no longer exist. You may have to search for some old provincial maps of the former Republic of Vietnam in order to understand where some of these provinces are/were (especially if you're interested in the regional variation noted by Paige (1970)).
Los Angeles Times maps inform the government control variable, and there are assumptions that Mitchell makes about the nature of control by the government (South Vietnam), the Vietcong, or the areas that are contested. The "control" here ultimately refers to South Vietnam.
The observations for government control variable are from 1965. Mitchell's footnote in his Table 1 says all other variables (except for population density) correspond with information from 1960. The population density estimate comes from 1964.
The coefficient of variation variable is defined as the standard deviation of land-holding size divided over the mean. If every landholding is of equal size, the observation is 0. Larger values suggest more variability in size of land-holdings with the implication being larger land-holdings are conspicuous in the province. It's a crude, but interesting, measure of inequality with that in mind.
The owner-operated land variable is another crude, but interesting/novel measure. An obvious percentage, 100 implies complete land ownership. 0 implies universal tenancy where peasants work on land they do not own. Some familiarity with the peculiarities of South Vietnamese society at the time is strongly suggested.
The "French land" and "Vietnamese land" variables refer to a specific agrarian reform measure ("Ordinance 57"). The Vietnam version includes both expropriated and redistributed land. The French version includes just expropriated land, per Mitchell. The logic is the Vietnamese version suggests higher values = lower inequality since the measure (partly) includes redistributed land. The French land, being just expropriation, has a single owner (the South Vietnamese government). That suggests higher inequality for higher values. This logic is interesting but questionable, and we'll just have to roll with the premise for the nature of the intended use of these data (i.e. replication). Paige's (1970) objection is more about regional variation in South Vietnam and its varied patterns of land use, and not about the particulars of these two measure (per se).
The mobility measure is a percentage, referring to the percentage of the province that is composed of plains and hills without dense forest.
The data are faithfully (to my level best!) scraped from Table 1 of his article. However, the results that come from a linear model do not perfectly reproduce his results (Equation 2, p. 432). I don't know why this is the case, nor is it that important. It is worth noting that this kind of "step-wise" procedure he employs for selecting a linear model is 100% not how you should do it, and that 33rd footnote he includes on p. 432 would be an automatic rejection at any quantitatively-oriented journal today.
It may interest the user to see re-analyses of Mitchell (1968) from around this time. I include those in the references for your consideration. Briefly, Paige's (1970) objection is that Mitchell (1968) includes radically different land-holding types into assorted measures of inequality and that Mitchell is selecting on 1965 (a watershed moment of insurgency during the war). Paranzino's (1972) critique is primarily statistical, though incorporates some of the issues raised by Paige (1970). Importantly, he correctly notes what the results of the linear model should be (p. 567).
Mitchell, Edward J. 1968. "Inequality and Insurgency: A Statistical Study of South Vietnam." World Politics 20(3): 421–38.
Paige, Jeffery M. 1970. "Inequality and Insurgency in Vietnam: A Re-Analysis." World Politics 23(1): 24–37.
Paranzino, Dennis. 1972. "Inequality and Insurgency in Vietnam: A Further Re-Analysis." World Politics 24(2): 565–78.
These are data you can use to replicate the regression discontinuity design analyses throughout Chapter 4 of Mastering 'Metrics. Original analyses come from Carpenter and Dobkin (2009, 2011).
mm_mldamm_mlda
A data frame with 50 observations on the following 19 variables.
agecella numeric
alla numeric
allfitteda numeric
internala numeric
internalfitteda numeric
externala numeric
externalfitteda numeric
alcohola numeric
alcoholfitteda numeric
homicidea numeric
homicidefitteda numeric
suicidea numeric
suicidefitteda numeric
mvaa numeric
mvafitteda numeric
drugsa numeric
drugsfitteda numeric
externalothera numeric
externalotherfitteda numeric
These data are not well-documented. You guys are on your own here. Good luck.
Carpenter, Christopher and Carlos Dobkin. 2009. "The Effect of Alcohol Consumption on Mortality: Regression Discontinuity Evidence from the Minimum Drinking Age". American Economic Journal: Applied Economics 1(1): 164–182.
Carpenter, Christopher and Carloss Dobkin. 2011. "The Minimum Legal Drinking Age and Public Health". Journal of Economic Perspectives 25(2): 133–156.
These are data from the 2009 NHIS survey. People who have read Mastering 'Metrics should recognize these data. They're featured prominently in that book and the authors' discussion of random assignment and experiments.
mm_nhismm_nhis
A data frame with 18790 observations on the following 10 variables.
fmlis the respondent a woman?
hia numeric vector for whether respondent has at least some health insurance
hltha numeric vector for a health index, broadly understood
nwhiteis the respondent not white?
agethe respondent's age in years
yeduthe respondent's total years of education
famsizethe size of the respondent's family
emplis the respondent employed
incthe respondent's household/family income
perweighta numeric vector for weight
Data are already cleaned in a way that facilitates an easy replication of Table 1.1 in Mastering 'Metrics. Check the book's website for more information.
National Health Interview Survey (2009).
These are data from the RAND Health Insurance Experiment (HIE). People who have read Mastering 'Metrics should recognize these data. They're featured prominently in that book and the authors' discussion of random assignment and experiments.
mm_randhiemm_randhie
The data are a list of two data frames (or "tibbles"). The first is the baseline data.
plantypethe plan coverage of the respondent, as a factor
agethe age of the respondent
blackhispwhether the respondent is not white
cholestthe cholesterol level of the respondent (in mg/dl)
educperthe education-level of the respondent
femalewhether the respondent is a woman
ghindxa general health index
hospwas the respondent hospitalized last year?
income1cpithe family/household income of the respondent, adjusted for inflation
mhia mental health index
systolthe systolic blood pressure level of the respondent (in mm HG)
The second is the outcome data.
plantypethe plan coverage of the respondent, as a factor
ftfthe number of face-to-face visits for the respondent
out_infthe total of out-patient expenses for the respondent
totadmthe number of hospital admissions for the respondent
tot_infthe total health expenses for the respondent
Data are already cleaned in a way that facilitates an easy replication of Table 1.3 and a partial replication of Table 1.4 in Mastering 'Metrics. Check the book's website for more information. I want to note that my treatment of the data leans heavily on Jeff Arnold's treatment of it. Check https://jrnold.github.io/masteringmetrics/ for more information. Future updates to the data may pursue a more exhaustive replication. I will only note these data are a mess and the authors of Mastering 'Metrics do not do a great job annotating code.
RAND Health Insurance Experiment.
Data, largely from Organisation Internationale des Constructeurs d'Automobiles (OICA), on motor vehicle production in various countries (and the world totals) from 1950 to 2019 at various intervals. Tallies include production of passenger cars, light commercial vehicles, minibuses, trucks, buses and coaches.
mvprodmvprod
A data frame with three variables
countrythe country's name
yearthe year
valuethe total motor vehicles produced that year
This is a Wikipedia web-scraping job. See: https://en.wikipedia.org/wiki/List_of_countries_by_motor_vehicle_production
Organisation Internationale des Constructeurs d'Automobiles (OICA)
This toy data set is loosely modified from Wave I of the NESARC data set. Here, my main interest is the number of drinks consumed on a usual day drinking alcohol in the past 12 months, according to respondents in the nationally representative survey of 43,093 Americans.
nesarc_drinkspdnesarc_drinkspd
A data frame with 43093 observations on the following 8 variables.
idnuma numeric vector and sequence from 1 to the number of rows in the data
ethrace2aa numeric vector for the ethnicity/race. 1 = White, not Hispanic. 2 = Black, not Hispanic. 3 = AI/AN. 4 = Asian, Native Hawaiian, Pacific Islander. 5 = Hispanic or Latino.
regiona numeric vector for the Census region. 1 = Northeast. 2 = Midwest. 3 = South. 4 = West
agea numeric vector for age in years
sexa numeric vector for sex. 1 = female. 0 = male
maritala numeric vector for marital status. 1 = married. 2 = living with someone as married. 3 = widowed. 4 = divorced. 5 = separated. 6 = never married
educa numeric vector for education level, recoded from s1q6a in the original data. 1 = did not make it to/finish high school. 2 = high school graduate or equivalency. 3 = some college, but no four-year degree. 4 = four-year college degree or more.
s2aq8ba numeric vector for the number of drinks of any alcohol consumed on days drinking alcohol in the past 12 months. This variable is “as-is” from the original data set.
You will not want to use the s2aq8b variable without recoding it first.
Those who cannot recall how much they typically drink (i.e. true don't knows'' or missing info) are coded as 99. Non-drinkers are coded as \code{NA} in the \code{s2aq8b} variable and should be recoded as 0. Any value between 1 and 98 in the variable represents the, for lack of better term, true” number of alcoholic drinks a respondent says s/he typically
consumes on a day drinking alcohol in the past 12 months, though this is evidently preposterous
as a count variable. A person drinking 42 alcoholic drinks a day would not be alive to tell you
they did this. The researcher may want to employ some sensible right censoring here.
National Epidemiologic Survey on Alcohol and Related Conditions (NESARC)—Wave 1 (2001–2002)
These are the data in Newhouse's (1977) simple OLS model from 1977. In his case, he's trying to explain medical care expenditures as a function of GDP per capita for these countries. It's probably the easiest OLS model I can find in print because Newhouse helpfully provides all the data in one simple table.
Newhouse77Newhouse77
A data frame with 13 observations on the following 5 variables.
countrya character vector for the country
yeara numeric vector for the year
gdppca numeric vector for the per capita GDP in USD
medsharegdpa numeric vector for the medical care share as percentage of GDP
medexppca numeric vector for per capita medical care expenditure (in USD)
Table 1 in Newhouse (1977) is well-annotated with background information.
Newhouse, Joseph P. 1977. "Medical-Care Expenditure: A Cross-National Survey." Journal of Human Resources 12(1): 115-125.
The NOAA Earth System Research Laboratory has an "ozone depleting gas index" (ODGI) data set from 1992 to 2018. This dataset summarizes Table 1 and Table 2 from its website. The primary interest here (for my purposes) is the ODGI indices (including the new 2012 measure). The data set includes constituent greenhouse gases/chlorines as well in parts per trillion. The primary use here is for in-class illustration.
ODGIODGI
A data frame with 62 observations on the following 16 variables.
yearthe year
catcategorical variable for the Antarctic or Mid-Latitudes measurements
cfc12CFC-12 concentration in parts per trillion
cfc11CFC-11 concentration in parts per trillion
ch3clchloromethane concentration in parts per trillion
ch3brbromomethane concentration in parts per trillion
ccl4carbon tetrachloride concentration in parts per trillion
ch3ccl3methyl chloroform concentration in parts per trillion
halonsaggregate concentration in parts per trillion of H-1211, H-1301 and H-2402
cfc113trichlorotrifluoroethane concentration in parts per trillion
hcfcsaggregate concentration in parts per trillion of HCFC-22, HCFC-141b, and HCFC-142b
wmo_minoraggregate concentration in parts per trillion of CFC-114, CFC-115, halon 2402 and halon 1201
sumthe sum of all greenhouse gas concentration measurements
eescincludes consideration of lag times for transport and mixing associated with transport. New as of 2012
odgi_oldold greenhouse gas index, no longer supported as of 2012
odgi_newnew greenhouse gas index, as of 2012
A data set for replicating an argument about the relationship between democracy and tariffs/non-tariff trade barriers.
OODTPTOODTPT
A data frame with 75 observations on the following 16 variables.
countrya character vector for the country
isocodea character vector for the three-character ISO code of the country
tariffthe mean statutory most favored nation tariff rate
corecovthe core non-tariff barrier coverage ratio
qualcovthe quality non-tariff barrier coverate ratio
politythe familiar Polity measure of democracy, from -10 to 10
iecthe index of electoral competitiveness from the World Bank
lngdppcreal GDP per capita in 1995 dollars
lngdpreal GDP in 1995 dollars
lnexpgdpexport dependence (i.e. export/GDP ratio)
reerreal effective exchange rate
growthGDP per capita growth rate
dimpgdpthe change in the import/GDP ratio over the past three years
lngovconsthe log of country's government consumption spending as a percentage of GDP
gatta dummy variable for GATT membership
avgtarthe country's average most favored nation tariff rate
Data downloaded Joshua Alley's Github repository on simple cross-sectional OLS models. These were originally two separate Stata files that I merged into one. Please read the Kono (2006) article for more information.
Kono, Daniel. 2006. "Optimal Obfuscation: Democracy and Trade Policy Transparency" American Political Science Review 100(3): 369-384.
A data set on the economic determinants of political unrest, for replicating a publication from 1973.
Parvin73Parvin73
A data frame with 26 observations on the following 9 variables.
countrya character vector for a country name
levviola numeric vector for the level of violence
pcia numeric vector for per capita income
incdista numeric vector for income distribution
d_pcia numeric vector for per capita income growth
sema numeric vector for socioeconomic mobility
cominta numeric vector for communication intensity
concfaca numeric vector for concentration factor
popa numeric vector for population size
The bulk of these data come from Russett's (1964) World Handbook of Political and Social Indicators. The data themselves are transcribed from the appendix of the article, which allows a replication of the results that Parvin (1973) reports. You should read that article for more information as to what's happening and for what purpose.
I did not catch Parvin (1973) mentioning this in the article, but there must be some kind of additive constant in the level of violence variable because the logarithmic transformations he reports would be undefined for the observations (like Denmark) where the level of violence is zero. The easiest way to approximate whatever Parvin (1973) did is to add .001 to the level of violence variable before taking its logarithmic transformation. That would allow a near perfect replication of Table 1.
It should go without saying that the population reported for Belgium, in the appendix, is likely a transcription error. Belgium's population is reported here as 9184, not "91.84.00".
The United Arab Republic was the short-lived union of Egypt and Syria, if you were curious what that is in the data.
Parvin, Manoucher. 1973. "Economic Determinants of Political Unrest: An Econometric Approach". Journal of Conflict Resolution 17(2): 271–96.
A data set on government spending in select rich countries as a function of trade/GDP, financial openness, and the state-year-level engagement in trade unions. The data offer a means to quasi-replicate Garrett's (1998) argument about left-wing governments' ability to stem the tide of globalization's effect on decreased government spending.
PPGEPPGE
A data frame with the following 9 variables.
countrya character vector for the country
iso3ca character vector for the three-character country ISO code
yearthe year
govtspendgdptotal government spending over GDP
tradegdpthe volume of trade over GDP
kaopenan index measuring a country's degree of capital account openness
ka_openan alternate index measuring a country's degree of capital account openness, normalized to be between 0 and 1
v2catraunian estimate of a country's engagement in independent trade unions, generated by way of a Bayesian item response model
v2catrauni_ordan estimate of a country's engagement in independent trade unions, on ordinal scale. See details.
The data are an unbalanced panel because of data missingness primarily affecting Switzerland (which would only appear in the panel in earnest starting in the mid-1990s). The Netherlands has some missing data in the mid-1970s. Spain and Portugal appear at the start of the panel, though the transition to democracy for both wouldn't start until 1974/1975. The data also have some obvious COVID weirdness for 2020. Perhaps an honest re-assessment of Garrett (1998) may want to drop Switzerland altogether, ignore 2020, and lop a few years off the Spanish and Portuguese panel. What you do with the Netherlands is up to you.
Briefly: the government spending/GDP data come from the International Monetary Fund. The trade/GDP data come from the World Bank's API. The financial openness indicators come by way of the Chinn-Ito index. The engagement in trade unions data are from the Varieties of Democracy project. The ordinal measure of the trade union estimates communicate what percentage of the population is active in independent trade unions. Values include 0) virtually no one 1) a small share of the population (less than 5%), 2) A moderate share of the population (about 5 to 15%). 3) A large share of the population (about 16 % to 25%). 4) A very large share of the population (about 26% or more).
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Nazifa Alizada, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Sandra Grahn, Allen Hicken, Garry Hindle, Nina Ilchenko, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Pamela Paxton, Daniel Pemstein, Josefine Pernes, Oskar Rydén, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Eitan Tzelgov, Yi-ting Wang, Tore Wig, Steven Wilson and Daniel Ziblatt. 2022. "V-Dem Country-Year/Country-Date Dataset v12" Varieties of Democracy (V-Dem) Project. doi:10.23696/vdemds22
Chinn, Menzie D. and Hiro Ito. 2006. "What Matters for Financial Development? Capital Controls, Institutions, and Interactions." Journal of Development Economics 81(1): 163–192.
Garrett, Geoffrey. 1998. Partisan Politics in the Global Economy New York, NY: Cambridge University Press.
A data set for replicating David Leblang's (1996) analysis on property rights, democracy, and economic growth.
PRDEGPRDEG
A data frame with 147 observations on the following 10 variables.
levinea numeric vector that servies as a cross-section identifier
countrya character vector for the country
decadea numeric vector for a decade
privatea numeric vector for credit allocated to private enterprise
rgdpa numeric vector for the initial level of real per capita GDP
democa numeric vector for the level of democracy
pria numeric vector for primary school attainment
seca numeric vector for secondary school attainment
growa numeric vector for per capita growth rate
xcontrola numeric vector for exchange controls
Data come Joshua Alley's Github repository on cross-sectional OLS regressions. Please read David Leblang's (1996) article for some more detail about the variables included in the model.
Leblang, David. 1996. "Property Rights, Democracy, and Economic Growth." 49(1): 5-26.
This should be self-evident. Here are all U.S. presidents who have completed their terms in office (i.e. excluding the current one).
PresidentsPresidents
A data frame with 45 observations on the following 3 variables.
presidentthe president
startthe start date of the term, as a date
endthe end date of the term, as a date
I scraped this from https://www.presidentsusa.net/presvplist.html.
Data frame is capital-P "Presidents" to avoid a conflict with the
presidents data frame from the datasets package.
These are some macroeconomic data for 21 select (rich) countries. I've used these data before to discuss issues of grouping and skew in cross-sectional data.
pwt_samplepwt_sample
A data frame with 1470 observations on the following 11 variables.
countrythe country name
isocodeThe country's ISO code
yeara numeric vector for the year
popPopulation in millions
hcIndex of human capital per person, based on years of schooling and returns to education
rgdpnaReal GDP at constant 2011 national prices (in million 2017 USD)
rgdpoOutput-side real GDP at chained PPPs (in million 2017 USD)
rgdpeExpenditure-side real GDP at chained PPPs (in million 2017 USD)
labshShare of labor compensation in GDP at current national prices
avhAverage annual hours worked by persons engaged.
empNumber of persons engaged (in millions)
rnnaCapital stock at constant 2017 national prices (in million 2017 USD)
Taken from the pwt10 package. See: https://www.rug.nl/ggdc/
These are four x-y data sets, combined into a long format, which have the same traditional statistical properties (mean, variance, correlation, regression line, etc.). However, they look quite different.
quartetsquartets
A data frame with 44 observations on the following 3 variables.
groupa categorical identifier for the quartet
xa continuous variable
ya continuous variable
Data come default in R, but I elected to change the format to be a bit more accessible.
Anscombe, Francis J. (1973). "Graphs in Statistical Analysis." The American Statistician 27: 17–21.
Data on U.S. recessions, past to present. Data include information on contraction, expansion, and cycle.
recessionsrecessions
A data frame with 35 observations on the following 8 variables.
peakthe year-month of the peak, as a date
troughthe year-month of the trough, as a date
peakqthe peak quarter
troughqthe trough quarter
p2tpeak to trough (in months)
prev_t2pprevious trough to this peak (in months)
tfpttrough from previous trough (in months)
pfpppeak from previous peak (in months)
Data come from via scraping job of https://www.nber.org/research/data/us-business-cycle-expansions-and-contractions
National Bureau of Economic Research (NBER)
A data set on banking, currency, debt, and debt-restructuring crises from 1970 to 2017.
SBCDSBCD
A data frame with 574 observations on the following 4 variables.
countrythe country, as it appears in the data
typethe type of crisis, entered here as "banking", "currency", "debt", or "debtrestructuring"
yearthe year of the crisis
monththe month the crisis started, if known
Data are cobbled from the second and third sheets of the spreadsheet the authors provide. Country names are as entered in their spreadsheet. Liberia has an "NA" in the raw data for sovereign debt restructuring and I don't know why. I elect to keep it.
Laeven, Luc and Fabian Valencia. 2020. "Systemic Banking Crises Database II". IMF Economic Review 68: 307–361.
This is a simple data set for matching region codes to the names of territorial units in Sweden, at least recorded/cataloged by the Central Bureau of Statistics in Sweden.
scb_regionsscb_regions
A data frame with 312 observations on the following 2 variables.
regionan intuitive name for a territorial unit/"region" in Sweden
region_codean alpha-numeric code coinciding with the territorial unit/"region"
Data were manually derived from first gathering everything the Central Bureau of Statistics had to offer. Its intended use is alongside the pxweb package. May it allow for more focused uses of the package without having to rely on the interactive component to do all the heavy-lifting.
County-level data on vote share and various background/demographic information for the 2016 South Carolina GOP/Democratic primaries.
SCP16SCP16
A data frame with 46 observations on the following 15 variables.
countythe county
clintonHillary Clinton's county-level vote share in the 2016 party primary
sandersBernie Sanders' county-level vote share in the 2016 party primary
trumpDonald Trump's county-level vote share in the 2016 party primary
cruzTed Cruz' county-level vote share in the 2016 party primary
rubioMarco Rubio's county-level vote share in the 2016 party primary
percapincA county-level estimate for per capita income
medhouseincA county-level estimate for the median household income
medfamincA county-level estimate for the median family income
illiteracyAn estimate of the percent of the county lacking "basic" prose literacy skills
perblackPercentage of the county that is black
populationAn estimate of the county-level population
romneyshare2012Mitt Romney's vote share at the county-level from the 2012 general election
perhsgradPercentage of the county whose residents 25 years and older have at least a high school education
unemploymentUnemployment rate for the county for January 2016
The illiteracy estimate comes from a Department of Education report from 2003. The unemployment rate data come from the Bureau of Labor Statistics. A Github repository contains more information: https://github.com/svmiller/sc-primary-2016.
These data describe how sea level has changed over time, in both relative and absolute terms. Absolute sea level change refers to the height of the ocean surface regardless of whether nearby land is rising or falling.
sealevelssealevels
A data frame with 136 observations on the following 5 variables.
yearthe year
adjlevadjusted sea level (in inches)
lbthe lower bound of the estimate (in inches)
ubthe upper bound of the estimate (in inches)
adjlev_noaaNOAA's adjusted sea level (in inches)
Environmental Protection Agency ("Climate Change Indicators: Sea Level")
CSIRO (Commonwealth Scientific and Industrial Research Organisation). 2015 update to data originally published in: Church, J.A., and N.J. White. 2011. Sea-level rise from the late 19th to the early 21st century. Surv. Geophys. 32:585–602.
NOAA (National Oceanic and Atmospheric Administration). 2016. Laboratory for Satellite Altimetry: Sea level rise. Accessed June 2016.
This data set contains yearly observations by the Environmental Protection Agency on the concentration of sulfur dioxide in parts per billion, based on 32 sites. I use this for in-class illustration. Note that the national standard is 75 parts per billion. Data are the national trend.
so2concentrationsso2concentrations
A data frame with the following 4 variables.
yearthe year
valuethe mean concentration of sulfur dioxide in the air based on 32 trend sites, in parts per billion
ubthe lower bound of the value (10th percentile)
lbthe upper bound of the value (90th percentile)
Environmental Protection Agency ("Sulfur Dioxide Trends")
A data set on state performance in inter-state wars. This data is useful for evaluating Valentino et al.'s (2010) "Bear Any Burden" analysis using more current data.
states_warstates_war
A data frame with the following variables.
micnuma numeric for the confrontation code
ccodea numeric for the Correlates of War state code
stdatea character vector communicating participant start date. See details for more.
enddatea character vector communicating participant start date. See details for more.
mindura numeric vector communicating minimum duration in confrontation. See details for more.
maxdura numeric vector communicating minimum duration in confrontation. See details for more.
sideaa numeric vector communicating whether participant was on side that initiated confrontation
origa numeric vector communicating whether participant was in confrontation on day one
hiacta numeric vector communicating highest action during confrontation
fatalmina numeric vector for minimum estimated fatalities for participant
fatalmaxa numeric vector for maximum estimated fatalities for participant
oppfatalmina numeric vector for minimum estimated fatalities by participant against opponents
oppfatalmaxa numeric vector for maximum estimated fatalities by participant against opponents
milexan estimate of military expenditures (in thousands)
milperan estimate of the size of military personnel (in thousands) for the state
cincThe Composite Index of National Capability ("CINC") score
tpopan estimate of the total population size of the state (in thousands)
v2x_polyarchythe Varieties of Democracy "polyarchy" estimate
polity2the the polity2 score from the Polity project
xm_qudsestan extension of the Unified Democracy Scores (UDS) estimates, made possibly by the QuickUDS package from Xavier Marquez.
wbgdp2011esta numeric vector for the estimated natural log of GDP in 2011 USD (log-transformed)
wbpopesta numeric vector for the estimated population size (log-transformed)
wbgdppc2011esta numeric vector for the estimated GDP per capita (log-transformed)
Start date and end date are in "MM/D(D)/YYYY" format. You can extract this
information into multiple columns with a separate function from the
tidyr package. This is mostly for convenience. Be mindful of two things:
First, dates are dates of first and last action, and not necessarily the
escalation to war, per se. Second, dates can be "missing". These are -9s, and
are commonplace when archival research can't pinpoint an exact day something
happened.
Observations select at the confrontation-level where maximum fatalities are greater than 1,000 and at the participant-level where (1) the participant engaged in at least an attack during this confrontation, (2) there are no instances where a participant dropped in/out on the same side of a multilateral confrontation or switched sides, and (3) the confrontation doesn't have an instance where a participant incurred fatalities while themselves not initiating a use of force. For illustration's sake, the Taiwan Straits Crises saw several appearances by the United States, but only one instance (for six days in Feb. 1953) where the U.S. engaged in an attack. World War II is a classic case of participants switching sides (France did so three times), but it also happened in the War of Latvian Independence as well (MIC#2604). The War of Attrition also saw the Russians reappear twice. Cases like these aren't included, mostly for convenience sake. In total, 41 cases with 1,000 maximum fatalities or more at the confrontation-level are excluded because of this. Of these 41 cases, World War II and the Vietnam War are the most conspicuous by their absence. Data come from version 1.01 of the Militarized Interstate Confrontation data.
Opponent fatalities are strictly dyadic and are derived from the Militarized Interstate Events data.
Capabilities, GDP, and democracy data come from peacesciencer for a forthcoming v. 1.2.0 release. See package for more information, though references are also included below. Variables are mostly lagged to the year prior to the participant observation year. However, there are several cases in the data that are born into war (see: India, Pakistan, North and South Korea, North and South Vietnam). In cases of missing data, information from the observation year is used.
The tpop and wbpopest columns are measuring the same thing but
are derived from two different data sets with two different data-generating
procedures. Use whichever one you like, but be mindful of what you're doing
and for what purpose you're doing it.
Anders, Therese, Christopher J. Fariss, and Jonathan N. Markowitz. 2020. "Bread Before Guns or Butter: Introducing Surplus Domestic Product (SDP)" International Studies Quarterly 64(2): 392–405.
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, M. Steven Fish, Adam Glynn, Allen Hicken, Anna Luhrmann, Kyle L. Marquardt, Kelly McMann, Pamela Paxton, Daniel Pemstein, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Agnes Cornell, Lisa Gastaldi, Haakon Gjerlow, Valeriya Mechkova, Johannes von Romer, Aksel Sundtrom, Eitan Tzelgov, Luca Uberti, Yi-ting Wang, Tore Wig, and Daniel Ziblatt. 2020. "V-Dem Codebook v10" Varieties of Democracy (V-Dem) Project.
Gibler, Douglas M., and Steven V. Miller. Forthcoming. “The Militarized Interstate Events (MIE) Dataset, 1816–2014.” Conflict Management and Peace Science.
Gibler, Douglas M., and Steven V. Miller. 2023. “The Militarized Interstate Confrontation Dataset, 1816-2014.” Journal of Conflict Resolution 68(2–3): 562–86
Marshall, Monty G., Ted Robert Gurr, and Keith Jaggers. 2017. "Polity IV Project: Political Regime Characteristics and Transitions, 1800-2017." Center for Systemic Peace.
Marquez, Xavier, "A Quick Method for Extending the Unified Democracy Scores" (March 23, 2016). doi:10.2139/ssrn.2753830
Miller Steven V. 2022. “peacesciencer: An R Package for Quantitative Peace Science Research.” Conflict Management and Peace Science, 39(6), 755–779.
Pemstein, Daniel, Stephen Meserve, and James Melton. 2010. "Democratic Compromise: A Latent Variable Analysis of Ten Measures of Regime Type." Political Analysis 18(4): 426-449.
Singer, J. David, Stuart Bremer, and John Stuckey. (1972). "Capability Distribution, Uncertainty, and Major Power War, 1820-1965." in Bruce Russett (ed) Peace, War, and Numbers, Beverly Hills: Sage, 19-48.
Singer, J. David. 1987. "Reconstructing the Correlates of War Dataset on Material Capabilities of States, 1816-1985" International Interactions, 14: 115-32.
Valentino, Benjamin A., Paul K. Huth, and Sarah E. Croco. 2010. "Bear Any Burden? How Democracies Minimize the Costs of War." Journal of Politics 72(2): 528-544
I cobbled together this data set of the professional clothes (polos, long-sleeve dress shirts, pants) in my closet, largely for illustration on the origins of apparel in the U.S. for an intro lecture on trade.
steves_clothessteves_clothes
A data frame with 86 observations on the following 4 variables.
typeType of clothing
brandThe brand of clothing (e.g. Apt. 9, Saddlebred)
colorthe color (and/or pattern) of the article of clothing
originThe country that produced the garment.
If you must know, I do most of my clothes shopping at major retailers in the U.S. This is mostly Belk, J.C. Penney, and Kohl's. If that's you as well, the odds are good the distribution of my clothes will closely resemble yours. A recent move I made resulted in me donating a fair bit of my short-sleeved polo shirts. I did not buy any new shirts, though. Thus, I copied that information from a previous version of the data.
Steve's closet. Hey, that's me!
This is primary commodity price data for sugar globally, in the United States, and in Europe for every month from 1980 to (roughly) the present. Prices are nominal U.S. cents per pound and are not seasonally adjusted ("NSA").
sugar_pricesugar_price
A data frame with 1,316 observations on the following 3 variables.
datea date
categorythe category (either the U.S., global, or Europe)
valuethe price of sugar in U.S. cents per pound (NSA, nominal)
The price data for Europe do not appear to be updated as regularly as the global and U.S. prices. Thus, the last month in the data for Europe are June 2017. For that reason, I elected to make a data set of these data for posterity's sake.
International Monetary Fund
A simple data set on Sweden's counties.
sweden_countiessweden_counties
A data frame with 21 observations on the following 6 variables.
isothe ISO 3166-2 code for the county
nutsthe Nomenclature of Territorial Units for Statistics (NUTS) code for the county
countythe name of the county, in Swedish
centrethe administrative centre, or centres, of the county
areathe size of the county in square kilometers
pop2019the size of the county in 2019
This is a simple Wikipedia scrape job from 7 November 2022.
A data set on satisfaction/dissatisfaction ratings during Margaret Thatcher's tenure as prime minister.
thatcher_approvalthatcher_approval
A data frame with 125 observations on the following 8 variables.
poll_datethe effective "date" of the public opinion poll
datea date for the poll, to make for easier plotting
govt_satthe percentage of respondents saying they were satisfied with the government
govt_disthe percentage of respondents saying they were dissatisfied with the government
thatcher_satthe percentage of respondents saying they were satisfied with Margaret Thatcher
thatcher_disthe percentage of respondents saying they were dissatisfied with Margaret Thatcher
opp_satthe percentage of respondents saying they were satisfied with the leader of the opposition
opp_disthe percentage of respondents saying they were dissatisfied with the leader of the opposition
Data come from Ipsos. "Leader of the opposition" was typically named in the exact poll. In the lifetime of this series, the leader of the opposition was James Callaghan until Nov. 10 1980. Thereafter, it was Michael Foot until Oct. 2 1983. Neil Kinnock replaces him for the duration of this series. Interpret "leader of the opposition" with that in mind.
The date variable is, again, for simple convenience to make for easier plotting. In the absence of a specific day provided by Ipsos, the poll benchmarks to the first of the month. In the case of a known period of days, it benchmarks to the first day provided.
A data set on thermometer ratings for Donald Trump and Barack Obama in 2020. I use these data for in-class illustration of central limit theorem. Basically: the sampling distribution of a population is normal, even if the underlying population is decidedly not.
thermstherms
A data frame with 3080 observations on the following 2 variables.
fttrump1a thermometer rating for Donald Trump
ftobama1a thermometer rating for Barack Obama
The survey period was April 10-18, 2020 and was done entirely online. Thermometer ratings are on a 0 to 100 scale, where higher values indicate more "warmth".
American National Election Studies (ANES) Exploratory Testing Survey (ETS)
A data set on turnip prices from my experience with Animal Crossing (New Horizons)
turnipsturnips
A data frame with the following 3 variables.
datea date
timea character vector referring to the particular time period of observation
pricea numeric vector for the price of turnips, in bells
Sunday prices are set for purchase and do not fluctuate. Timmy and Tommy do not accept turnips on Sunday either. Daily prices fluctuate both at opening on Nook's Cranny and at noon. This amounts to three time periods in the data. "5:00 a.m." is reserved only for Sunday purchases (i.e. when Daisy Mae arrives on the island). 8:00 a.m. is the morning price because that is when Nook's Cranny opens. 12:00 p.m. is when the price changes for the day.
Explanations for missing dates: Timmy and Tommy were renovating the shop on May 6, 2021. My wife was diagnosed with cancer and my mother in law went to the hospital on the afternoon of Dec. 27, 2021. I did not get to play the game on Jan. 9, 2022 because of errands I was running for my wife. I plain forgot to check on Feb. 7, 2022.
These data come from the 2016 CCES and allow interested students to model the individual correlates of the Trump vote in 2016. Code/analysis heavily indebted to a 2017 analysis I did on my blog (see references).
TV16TV16
A data frame with 64600 observations on the following 21 variables.
uida numeric vector, a unique identifier for the respondent as they first appear in the CCES data.
statea character vector for the state in which the respondent resides
votetrumpa numeric that equals 1 if the respondent voted says s/he voted for Trump in 2016.
agea numeric vector for age that is roughly calculated as 2016 - birthyr, as it's coded in the CCES data.
femalea numeric that equals 1 if the respondent is a woman
collegeeda numeric vector that equals 1 if the respondent says s/he has a college degree
racefa character vector for the race of the respondent
famincra numeric vector for the respondent's household income. Ranges from 1 (Less than $10,000) to 12 ($150,000 or more).
ideoa numeric vector for the respondent's ideology on a liberal-conservative discrete scale. 1 = very liberal. 5 = very conservative.
pid7naa numeric vector for the respondent's partisanship on the familiar 1-7 scale. 1 = Strong Democrat. 7 = Strong Republican. Other party supporters (e.g. libertarians) are coded as NA.
bornagaina numeric vector for whether the respondent self-identifies as a born-again Christian.
religimpa numeric vector for the importance of religion to the respondent. 1 = not at all important. 4 = very important.
churchatda numeric vector for the extent of church attendance for the respondent. 1 = never. 6 = more than once a week.
prayerfreqa numeric vector for the frequency of prayer for the respondent. 1 = never. 7 = several times a day.
angryracisma numeric vector for how angry the respondent is that racism exists. 1 = strongly agree (i.e. is angry racism exists). 5 = strongly disagree.
whiteadva numeric vector for agreement with statement that white people have advantages over others in the U.S. 1 = strongly agree. 5 = strongly disagree.
fearracesa numeric vector for agreement with statement that the respondent fears other races. 1 = strongly disagree. 5 = strongly agree.
racerarea numeric vector for agreement with statement that racism is rare in the U.S. 1 = strongly disagree. 5 = strongly agree.
lreliga numeric vector that serves as a latent estimate for religiosity from the bornagain, religimp, churchatd, and prayerfreq variables. Higher values = more religiosity.
lcograca numeric vector that serves as a latent estimate for cognitive racism. This is derived from the racerare and whiteadv variables.
lempraca numeric vector that serves as a latent estimate for empathetic racism. This is derived from the fearraces and angryracism variables.
The latent estimates for religiosity, cognitive racism, and empathetic
racism come from a graded response model estimated in mirt. The concepts of
"cognitive racism" and "empathetic racism" come from DeSante and Smith.
Cooperative Congressional Election Study, 2016
http://svmiller.com/blog/2017/04/age-income-racism-partisanship-trump-vote-2016/
https://github.com/svmiller/2016-cces-trump-vote/blob/master/1-2016-cces-trump.R
This is a (near) daily data set on the effective exchange rate index for the United Kingdom's pound sterling from 1990 onward. The data are indexed, such that 100 equals the monthly average in January 2005. This is useful for illustrating devaluations of the pound after Black Wednesday, the financial crisis, and, more recently, the UK's separation from the European Union.
ukg_eeriukg_eeri
A data frame with 8318 observations on the following 2 variables.
datea date
valuea numeric vector for the effective exchange rate index (Jan. 2005 = 100)
Credit to the Bank of England for making these data readily available and
accessible. The Bank of England's website (https://www.bankofengland.co.uk/) has
these data with a code of XUDLBK67.
Bank of England
Cross-national data on relative size of the trade unions and predictors in 20 countries. This is a data set of interest to replicating Western and Jackman (1994), who themselves were addressing a debate between Wallerstein and Stephens on which of two highly correlated predictors explains trade union density.
uniondensityuniondensity
A data frame with 20 observations on the following 5 variables.
countrya character vector for the country
uniona numeric vector for the percentage of the total number of wage and salary earners plus the unemployed who are union members, measured between 1975 and 1980, with most of the data drawn from 1979.
lefta numeric vector tapping the extent to which parties of the left have controlled governments since 1919, due to Wilensky (1981).
sizea numeric vector measuring the log of labor force size, defined as the number of wage and salary earners, plus the unemployed.
concena numeric vector measuring the percentage of employment, shipments, or production accounted for by the four largest enterprises in a particular industry, averaged over industries (with weights proportional to the size of the industry) and the resulting measure is normalized such that the United States scores a 1.0, and is due to Pryor (1973). Some of the scores on this variable are imputed using procedures described in Stephens and Wallerstein (1991, 945).
Data documentation are derived from Simon Jackman's pscl package.
I just tidied up the presentation a bit.
Pryor, Frederic. 1973. Property and Industrial Organization in Communist and Capitalist Countries. Bloomington: Indiana University Press.
Stephens, John and Michael Wallerstein. 1991. Industrial Concentration, Country Size and Trade Union Membership. American Political Science Review 85:941-953.
Western, Bruce and Simon Jackman. 1994. Bayesian Inference for Comparative Research. American Political Science Review 88:412-423.
Wilensky, Harold L. 1981. Leftism, Catholicism, Democratic Corporatism: The Role of Political Parties in Recemt Welfare State Development. In The Development of Welfare States in Europe and America, ed. Peter Flora and Arnold J. Heidenheimer. New Brunswick: Transaction Books.
Jackman, Simon. 2009. Bayesian Analysis for the Social Sciences. Wiley: Hoboken, New Jersey.
This is a toy data set to examine the time in which we should expect China
to overtake the United States in total gross domestic product (GDP),
given current trends. It includes an OECD long-term GDP forecast from 2014,
and forecasts from the forecast and prophet packages in R.
usa_chn_gdp_forecastsusa_chn_gdp_forecasts
A data frame with 182 observations on the following 12 variables.
countrya character vector (United States, China)
yeara numeric vector for the year
p_gdpy-hats (forecasted GDP) from a prophet forecast
p_lo80lower bound (80%) of y-hats (forecasted GDP) from a prophet forecast
p_hi80upper bound (80%) of y-hats (forecasted GDP) from a prophet forecast
gdpobserved GDP, made available to the World Bank and OECD national accounts data. Available from 1960 to 2019.
f_gdpforecasted GDP from 2020 to 2050, from the forecast package
f_lo80lower bound (80%) forecasted GDP from 2018 to 2050, from the forecast package
f_hi80upper bound (80%) forecasted GDP from 2018 to 2050, from the forecast package
f_lo95lower bound (95%) forecasted GDP from 2018 to 2050, from the forecast package
f_hi95upper bound (95%) forecasted GDP from 2018 to 2050, from the forecast package
oecd_ltgdpflong-term GDP forecast from the OECD via the OECD Outlook No 95 - May 2014
Forecasts from the forecast package and prophet
package are rudimentary and bare minimum forecasts based on previous values
to that point. Notice the forecast forecasts have a prefix of
f_ and the prophet forecasts have a prefix of
p_. Forecasts are not meant to be exhaustive (clearly), only
illustrative for in-class discussion about the "Rise of China." Forecasts
made in R on Nov. 20, 2020.
OECD Outlook No 95 - May 2014 - Long-term baseline projections provided by Organisation for Economic Co-operation and Development (OECD)
This is a simple and regrettably incomplete time-series on the percentage of U.S. households with access to a computer, by year.
usa_computersusa_computers
A data frame with 19 observations on the following 2 variables.
yearthe year
valuethe estimated percentage of households with access to a computer
Data are spotty and regrettably this is not a perfect time-series. However, it is useful for an in-class exercise to show that the proliferation of household computers (over time) in the United States comes in part because of globalization. Use it for that purpose. The data are reasonably faithful, but don't treat it as gospel. Exact sourcing available upon request.
Various: U.S. Census Bureau, Current Population Survey, and American Community Survey
This data set contains counts/estimates for the number of inbound migrants in the U.S as well as outbound migrants of American origin to other countries from 1990 to 2017.
usa_migrationusa_migration
A data frame with 3535 observations on the following 5 variables.
yeara numeric vector for 1990, 1995, 2000, 2005, 2010, 2015, 2017
countrya character vector/constant for the United States
categorya character vector for whether the count is inbound to the U.S. from the area variable or outbound (i.e. American expats) to the area variable in a given year.
areaa character vector for the area of origin (if category == "Inbound") or destination for American migrants (if category == "Outbound")
counta numeric vector for the count of inbound/outbound migrants
"Cote d'Ivoire", "Curacao", and "Reunion" originally had UTF-8 characters, which were removed for maximal compliance with CRAN. CRAN raises a note for every non-ASCII character it sees.
United Nations Population Division (DESA)
A simple data set from state.abb, state.name, state.region, and
state.division (+ District of Columbia). I'd rather just have all these
in one place.
usa_statesusa_states
A data frame with 51 observations on the following 4 variables.
stateabbthe state abbreviation
statenamethe state's name
regionthe state's Census region
divisionthe state's Census division
A yearly data set on U.S. trade and GDP from 1790 to 2018. Data also include a population variable to facilitate per capita adjustments, if the user sees it useful.
usa_tradegdpusa_tradegdp
A data frame with 229 observations on the following 5 variables.
yearthe year
gdpbU.S. GDP (nominal, in billions)
popPopulation of the U.S. (in thousands)
impoThe value of U.S. imports (in billions)
expoThe value of U.S. exports (in billions)
Data come from various sources (see, especially: https://econdataus.com/tradeall.html). Post-1989 data come from the U.S. Census Bureau. 2018 GDP comes from the IMF. 2018 population estimate comes from the U.S. Census Bureau.
A data set on economic aid allocation by the United States for assorted years. These are useful for illustrative cross-sectional relationships between human rights and U.S. aid allocation at what amounts to midway points for various presidential administrations.
USFAHRUSFAHR
A data frame with 1654 observations on the following 18 variables.
countryan English country name
ccodea Correlates of War state code
regiona region in which the country resides, per Greenbook
yeara year
nomobligeconomic aid obligations in nominal U.S. dollars
constobligeconomic aid obligations in costant 2019 U.S. dollars
clphya physical violence index, bound between 0 and 1
civliba civil liberties index, bound between 0 and 1
fpsusaforeign policy similarity with the United States
fpsrusforeign policy similarity with the Soviet Union/Russia
mindistusaminimum distance of the country from the United States
mindistrusminimum distance of the country from the USSR/Russia
gdpan estimate of GDP in constant 2011 U.S. dollars
popan estimate of population size
usaimpa value of how much the U.S. imports from the country (in thousands USD)
usaexpa value of how much the U.S. exports to the country (in thousands USD)
milexan estimate of military expenditures (in thousands USD)
cinca composite index of national capabilities
Matching is done on Correlates of War state codes. Thus, the exact "population" is an amalgam of U.S. aid and Correlates of War state system membership. Regions are offered, as is, from USAID Data Services.
Data on aid are "obligations" and not "disbursements", and thus may better reflect donor intent. These come from US Overseas Loans & Grants ("Greenbook") and were prepared by USAID Data Services on July 14, 2021.
Greenbook only offers information about dollar amounts of aid, contingent on receiving aid. Observations were added, based on Correlates of War state system membership, about countries that could've received aid but did not. Countries that never received aid at all had to have regions assigned to them ex post. I don't think the regions imputed for these observations are problematic. This concerns Andorra, Czechoslovakia, Dominica, German Democratic Republic, German Federal Republic, Liechtenstein, Luxembourg, Monaco, Nauru, Republic of Vietnam, San Marino, St. Lucia, St. Kitts and Nevis, Switzerland, Tuvalu, Yemen Arab Republic, Yemen People's Republic, and Zanzibar.
Higher values of the physical violence index and civil liberties index communicate better human rights records. Data are lagged a year.
Foreign policy similarity is Cohen's (1960) kappa based on valued United
Nations General Assembly voting. Data come from Haege (2011) by way of
peacesciencer's add_fpsim() function. Please read
peacesciencer documentation for more information about these measures,
along with what you should cite for any serious use of these data. Higher
values for these measures = more foreign policy similarity.
Minimum distance is calculated using the Vincenty method ("as the crow
flies"). Measurement is in kilometers and data come from peacesciencer
and its add_minimum_distance() function. Check package documentation
for appropriate citation for any serious use.
Estimates of gross domestic product ("GDP") and population come by way of
peacesciencer and its add_sdp_gdp() function. Check package
documentation for appropriate citations for any serious use. GDP is in
actual dollars.
Trade data come from Correlates of War trade data by way of peacesciencer
and its add_cow_trade() function. Check package documentation for
appropriate citations for any serious use.
Military expenditure and capabilities data come from Correlates of War by way
of peacesciencer and its add_cow_trade() function. Check package
documentation for appropriate citations for any serious use.
A data set on turnout and demographic data from the 2000 Current Population Survey. This is a basic
port of the voteincome data from the Zelig package.
voteincomevoteincome
A data frame with 1500 observations on the following 7 variables.
statea character variable for the state, either Arkansas (AK) or South Carolina (SC)
yeara numeric constant for the year (2000)
votea dummy variable for whether the person voted (1) or did not vote
incomea numeric variable for income ranging from 4 (less than $5000) to 17 (greater than $75000)
educationa numeric variable for educational attainment ranging from 1 (less than high school education) to 4 (more than college education)
agea numeric variable for the respondent's age in years,ranging from 18 to 85
femalea dummy variable for whether the respondent is a woman (1) or a man (0)
Data come from the 2000 Current Population Survey by way of the Zelig package. Data should not be used for inferential applications, only for pedagogical purposes. See the appropriate CPS codebook for more information on variable coding (especially for income and education). In all likelihood, age is right-censored as well.
A simple data set drawn from World Bank Open Data. I'll use it to illustrate some merge issues you might encounter in panel data.
wbd_examplewbd_example
A data frame with 4537 observations on the following 7 variables.
countryan English name for the country/territorial unit
iso2cthe two-character ISO code for the country/territorial unit
iso3cthe three-character ISO code for the country/territorial unit
yearthe year of observation
rgdppcthe real GDP per capita of the country/territorial unit in that year
lifeexpthe average life expectancy at birth for men and women for the country that year
hcithe human capital index for the country that year
The idea for this data comes by way of a student encounter where we noticed this issue. Data were further generated by the wonderful WDI package. The underlying data come from the World Bank national accounts (GDP per capita), World Bank analyst estimates (human capital index), or the United Nations Population Division (life expectancy at birth).
The human capital index is on a 0 to 1 scale.
A simple data set that syncs World Values Survey country codes (s003)
with corresponding country codes from the Correlates of War state system
membership data.
wvs_ccodeswvs_ccodes
A data frame with 112 observations on the following 3 variables.
s003the World Values Survey country code
countrya character vector for the corresponding country name
ccodethe equivalent country code from the Correlates of War state system membership data
http://svmiller.com/blog/2015/06/syncing-word-values-survey-country-codes-with-cow-codes/
A data set on attitudes about immigration for all observations in the third to sixth wave of the World Values Survey. I use these data for in-class illustration.
wvs_immigwvs_immig
A data frame with 310,388 observations on the following 6 variables.
s002the World Values Survey wave
s003the World Values Survey country code
countrythe country name
s020the survey year
uida unique identifier for the survey respondent
e143an attitude about immigration policy in the World Values Survey
1 = "let anyone come". 2 = "as long as jobs are available". 3 = "strict limits".
4 = "Prohibit people from coming" for the e143 variable. See ?wvs_ccodes for
more information about naming/identifying countries.
A data set on attitudes about the justifiability of bribe-taking for all observations in the third to sixth wave of the World Values Survey. I use these data for in-class illustration about seemingly interval-level, but information-poor measurements.
wvs_justifbribewvs_justifbribe
A data frame with 348532 observations on the following 6 variables.
s002the World Values Survey wave
s003the World Values Survey country code
countrythe country name
s020the survey year
uida unique identifier for the survey respondent
f117an attitude about the justifiability of bribe-taking in the World Values Survey
1 = "never justifiable". 10 = "always justifiable". Increasing values on this 1-10 scale imply increasing permissiveness for the respondent toward this particular/blatant form of corruption.
A data set on attitudes about the justifiability of abortion in the United States based on World Values Survey responses recorded across six waves (from 1982 to 2011). I assembled this data frame probably around 2014 and routinely use it for in-class illustration about regression, post-estimation simulation, quantities of interest, and how to think about modeling a dependent variable that is on a 1-10 scale, but has curious heaping patterns.
wvs_usa_abortionwvs_usa_abortion
A data frame with 10387 observations on the following 16 variables.
wvsccodethe country code for the United States (a numeric constant)
wavethe survey wave
yearthe survey year corresponding to the survey wave
ajthe justifiability of abortion on a 1-10 scale (1 = never justifiable; 10 = always justifiable)
agethe age of the respondent in years
collegeeda dummy variable that equals 1 if the respondent graduated from college
femalea dummy variable that equals 1 if the respondent is a woman
unemployeda dummy variable that equals 1 if the respondent is unemployed
ideologythe ideological self-placement of the respondent on a 1-10 scale (1 = furthest to the left; 10 = furthest to the right)
satisfinancialthe respondent's financial satisfaction with his/her life (1 = most dissatisfied; 10 = most satisfied)
postma4the post-materialist index for the respondent (-1 = materialist; 0 = mixed, 1 = post-materialist)
caithe child autonomy index, which ranges from -2 to 2
trustmostpeoplecan most people be trusted (1) or "(you) never can be too careful" (0)
godimportantthe importance of God to the respondent on a 1-10 scale (1 = God is not at all important; 10 = God is most important)
respectauthoritywould more respect for authority be a welcome change to the United States?
nationalpridea dummy that equals 1 if the respondent is very proud to be an American.
Data come from the World Values Survey. Note that the college education variable
is curiously NA until the third survey wave. The child autonomy index ranges from
-2 to 2 where increasing values indicate that children should learn determination and
independence over obedience and religious faith. The respectauthority variable is coded
where -1 means the respondent believes greater respect for authority in the United States
as a future change to the country would be a bad thing. 0 means the respondent doesn't mind
such a change. 1 = the respondent believes it would be a good thing.
This is a simple data set that summarizes what the education codes are in the World Values Survey for the United States.
wvs_usa_educatwvs_usa_educat
A data frame with 42 observations the following 6 variables.
x025the numeric code for supposedly the highest educational level attained
x025cswvsthe numeric code for supposedly the education-level attained by the respondent, with country-specific categories
nthe number of observations in the World Values Survey with that unique x025cswvs code
x025cswvsmeaningthe meaning behind the unique x025cswvs code
x025meaningthe meaning behind the unique x025 code
educata standardized categorical variable corresponding with that unique x025cswvs code
Observations taken from the combined seven waves of survey data made available by the World Values Survey, but isolated to just the United States. The World Values Survey unfortunately did not collect information about the education-level of the respondent in the 1981 and 1990 waves. These education categories feature in the Miller and Davis (2020) article in Journal of Ethnicity, and Politics, albeit before the release of the seventh wave.
Miller, Steven V. and Nicholas T. Davis. Forthcoming. "The Effect of White Social Prejudice on Support for American Democracy." Journal of Race, Ethnicity, and Politics.
This is a simple data set that summarizes what the region codes are in the World Values Survey for the United States.
wvs_usa_regionswvs_usa_regions
A data frame with 63 observations the following 6 variables.
x048wvsthe numeric code for supposedly the region in which the interview was conducted
x048wvsmeaningthe meaning behind the unique x048wvs code
stateabbthe corresponding state abbreviation (if available) for the unique x048wvs code
statenamethe corresponding state abbreviation (if available) for the unique x048wvs code
divisionthe corresponding division for the unique x048wvs code
regionthe corresponding region for the unique x048wvs code
The region codes are a mess. Some of these are informed guesses. For example, I assume "Northwest" means "Pacific" and that Idaho was not included in that category. I make a similar assumption that "Rocky Mountain state" means "Mountain".
A data set on Yugo sales against two competing models in the United States from 1985 to 1992.
yugo_salesyugo_sales
A data frame with 24 observations on the following 3 variables.
yearthe year
carthe car type, either the Hyundai Excel, Yugo, or Toyota Tercel
salesthe number of units sold in the United States
Data come from a website then known as carsalesbase.com. I'm
aware the inclusion of the Tercel is questionable since the third generation
of Tercels were quite different from the first and second generations.
However, I use these data to illustrate how poorly the Yugo fared against
competing models, including the first and second generation Tercels. I think
the inclusion is fair for that purpose.