Burde and Linden (2013, AEJ Applied) Dataset

Description

Replicates IV using controls from Table 2

Usage

afghan

Format

A data frame with 687 rows and 17 variables:

enrolled

Indicator if child is enrolled in formal school. Outcome.

testscore

Normalized test score

buildschool

Indicator if village is treated. Instrument.

headchild

Indicator if child is child of head of household

nhh

Number of household members

female

Female indicator

age

Child's age

yrsvill

Time family has lived in village

farsi

Indicator for speaking Farsi

tajik

Indicator for speaking Tajik

farmers

Indicator for if head of household is a farmer

land

Number of jeribs of land owned

agehead

Head of household age

educhead

Years of education for head of household

sheep

Number of sheep and goats owned

chagcharan

Indicator if village is in Chagcharan district

distschool

Distance to nearest non-community based school

Source

Provided by author.

References

https://www.jstor.org/stable/3083335

B function from Proposition A3

Description

B function from Proposition A3

Usage

b_functionA3(obs_draws, g, psi)

Arguments

Value

A min and a max of the B function

Evaluates the corners given user bounds. Vectorized wrt multiple draws of obs.

Description

Evaluates the corners given user bounds. Vectorized wrt multiple draws of obs.

Usage

candidate1(r_TstarU_lower, r_TstarU_upper, k_lower, k_upper, obs)

Arguments

Value

List containing vector of lower bounds and vector of upper bounds of r_uz

Evaluates the edge where k is on the boundary. Vectorized wrt multiple draws of obs.

Description

Evaluates the edge where k is on the boundary. Vectorized wrt multiple draws of obs.

Usage

candidate2(r_TstarU_lower, r_TstarU_upper, k_lower, k_upper, obs)

Arguments

Value

List containing vector of lower bounds and vector of upper bounds of r_uz

Evaluates the edge where r_TstarU is on the boundary.

Description

Evaluates the edge where r_TstarU is on the boundary.

Usage

candidate3(r_TstarU_lower, r_TstarU_upper, k_lower, k_upper, obs)

Arguments

Value

List containing vector of lower bounds and vector of upper bounds of r_uz

Collapse 3-d array to matrix

Description

Collapse 3-d array to matrix

Usage

collapse_3d_array(myarray)

Arguments

Value

Matrix with the 3rd dimension appended as rows to the matrix

Acemoglu, Johnson, and Robinson (2001) Dataset

Description

Cross-country dataset used to construct Table 4 of Acemoglu, Johnson & Robinson (2001).

Usage

colonial

Format

A data frame with 64 rows and 9 variables:

shortnam

three letter country abbreviation, e.g. AUS for Australia

africa

dummy variable =1 if country is in Africa

lat_abst

absolute distance to equator (scaled between 0 and 1)

rich4

dummy variable, =1 for "Neo-Europes" (AUS, CAN, NZL, USA)

avexpr

Average protection against expropriation risk. Measures risk of government appropriation of foreign private investment on a scale from 0 (least risk) to 10 (most risk). Averaged over all years from 1985-1995.

logpgp95

Natural logarithm of per capita GDP in 1995 at purchasing power parity

logem4

Natural logarithm of European settler mortality

asia

dummy variable, =1 if country is in Asia

loghjypl

Natural logarithm of output per worker in 1988

Source

http://economics.mit.edu/faculty/acemoglu/data/ajr2001

References

https://www.aeaweb.org/articles.php?doi=10.1257/aer.91.5.1369

Computes bounds for simulated data

Description

This function takes data and user restrictions on measurement error and endogeneity and simulates data and the resulting bounds on instrument validity.

Usage

draw_bounds(
  y_name,
  T_name,
  z_name,
  data,
  controls = NULL,
  r_TstarU_restriction = NULL,
  k_restriction = NULL,
  n_draws = 5000
)

Arguments

Value

List containing simulated data observables (covariances, correlations, and R-squares), indications of whether the identified set is empty, the unrestricted and restricted bounds on instrumental relevance, instrumental validity, and measurement error.

Simulates different data draws

Description

This function takes the data and simulates potential draws of data from the properties of the observed data.

Usage

draw_observables(y_name, T_name, z_name, data, controls = NULL, n_draws = 5000)

Arguments

Value

Data frame containing covariances, correlations, and R-squares for each data simulation

Draws covariance matrix using the Jeffrey's Prior

Description

Draws covariance matrix using the Jeffrey's Prior

Usage

draw_sigma_jeffreys(y, Tobs, z, k, n_draws)

Arguments

Value

Array of covariance matrix draws

Creates LaTeX code for the HPDI

Description

Creates LaTeX code for the HPDI

Usage

format_HPDI(bounds)

Arguments

Value

LaTeX string of the HPDI

Creates LaTeX code for parameter estimates

Description

Creates LaTeX code for parameter estimates

Usage

format_est(est)

Arguments

Value

LaTeX string for the number

Creates LaTeX code for the standard error

Description

Creates LaTeX code for the standard error

Usage

format_se(se)

Arguments

Value

LaTeX string for the standard error

G function from Proposition A.2

Description

G function from Proposition A.2

Usage

g_functionA2(kappa, r_TstarU, obs_draws)

Arguments

Value

G value

Computes coverage of list of intervals

Description

Computes coverage of list of intervals

Usage

getCoverage(data, guess)

Arguments

Value

Coverage percentage

Generates smallest covering interval

Description

Generates smallest covering interval

Usage

getInterval(data, center, conf = 0.9, tol = 1e-06)

Arguments

Value

2-element vector of confidence interval

Computes L, lower bound for kappa_tilde in paper

Description

Computes L, lower bound for kappa_tilde in paper

Usage

get_L(draws)

Arguments

Value

Vector of L values

Solves for the magnification factor

Description

This function solves for the magnification factor given r_TstarU and kappa. It handles 3 potential cases when the magnification factor must be evaluated: 1. Across multiple simulations, but given the same r_TstarU and k 2. For multiple simulations, each with a value of r_TstarU and k 3. For one simulation across a grid of r_TstarU and k

Usage

get_M(r_TstarU, k, obs)

Arguments

Value

Vector of magnification factors

Computes a0 and a1 bounds

Description

Computes a0 and a1 bounds

Usage

get_alpha_bounds(draws, p)

Arguments

Value

List of alpha bounds

Solves for beta

Description

This function solves for beta given r_TstarU and kappa. It handles 3 potential cases when beta must be evaluated: 1. Across multiple simulations, but given the same r_TstarU and k 2. For multiple simulations, each with a value of r_TstarU and k 3. For one simulation across a grid of r_TstarU and k

Usage

get_beta(r_TstarU, k, obs)

Arguments

Value

Vector of betas

Returns beta bounds in binary case using grid search

Description

Returns beta bounds in binary case using grid search

Usage

get_beta_bounds_binary(obs_draws, p, r_TstarU_restriction)

Arguments

Value

Min and max values for beta

Generates beta bounds off of beta draws

Description

Generates beta bounds off of beta draws

Usage

get_beta_bounds_binary_post(draws, n_observables)

Arguments

Value

Upper and lower bounds of beta based on posterior draws

Wrapper function combines all unrestricted bounds together. Vectorized

Description

Wrapper function combines all unrestricted bounds together. Vectorized

Usage

get_bounds_unrest(obs)

Arguments

Value

List of unrestricted bounds for r_TstarU, r_uz, and kappa

Computes OLS and IV estimates

Description

Computes OLS and IV estimates

Usage

get_estimates(y_name, T_name, z_name, data, controls = NULL, robust = FALSE)

Arguments

Value

List of beta estimates and associated standard errors for OLS and IV estimation

Given observables from the data, generates unrestricted bounds for kappa. Vectorized

Description

Given observables from the data, generates unrestricted bounds for kappa. Vectorized

Usage

get_k_bounds_unrest(obs, tilde)

Arguments

Value

List of upper bounds and lower bounds for kappa

Computes beliefs that support valid instrument

Description

Computes beliefs that support valid instrument

Usage

get_new_draws(obs_draws, post_draws)

Arguments

Value

data.frame of new draws

Given data and function specification, returns the relevant correlations and covariances with any exogenous controls projected out.

Description

Given data and function specification, returns the relevant correlations and covariances with any exogenous controls projected out.

Usage

get_observables(y_name, T_name, z_name, data, controls = NULL)

Arguments

Value

List of correlations, covariances, and R^2 of first and second stage regressions after projecting out any exogenous control regressors

Compute the share of draws that could contain a valid instrument.

Description

Compute the share of draws that could contain a valid instrument.

Usage

get_p_valid(draws)

Arguments

Value

Numeric of the share of valid draws as determined by having the the restricted bounds for r_uz contain zero.

Computes the lower bound of psi for binary data

Description

Computes the lower bound of psi for binary data

Usage

get_psi_lower(s2_T, p, kappa)

Arguments

Value

Vector of lower bounds for psi

Computes the upper bound of psi for binary data

Description

Computes the upper bound of psi for binary data

Usage

get_psi_upper(s2_T, p, kappa)

Arguments

Value

Vector of upper bounds for psi

Given observables from the data, generates the unrestricted bounds for rho_TstarU. Data does not impose any restrictions on r_TstarU Vectorized

Description

Given observables from the data, generates the unrestricted bounds for rho_TstarU. Data does not impose any restrictions on r_TstarU Vectorized

Usage

get_r_TstarU_bounds_unrest(obs)

Arguments

Value

List of upper and lower bounds for r_TstarU

Solves for r_uz given observables, r_TstarU, and kappa

Description

This function solves for r_uz given r_TstarU and kappa. It handles 3 potential cases when r_uz must be evaluated: 1. Across multiple simulations, but given the same r_TstarU and k 2. For multiple simulations, each with a value of r_TstarU and k 3. For one simulation across a grid of r_TstarU and k

Usage

get_r_uz(r_TstarU, k, obs)

Arguments

Value

Vector of r_uz values.

Evaluates r_uz bounds given user restrictions on r_TstarU and kappa

Description

This function takes observables from the data and user beliefs over the extent of measurement error (kappa) and the direction of endogeneity (r_TstarU) to generate the implied bounds on instrument validity (r_uz)

Usage

get_r_uz_bounds(r_TstarU_lower, r_TstarU_upper, k_lower, k_upper, obs)

Arguments

Value

2-column data frame of lower and upper bounds of r_uz

Given observables from the data, generates the unrestricted bounds for rho_uz. Vectorized

Description

Given observables from the data, generates the unrestricted bounds for rho_uz. Vectorized

Usage

get_r_uz_bounds_unrest(obs)

Arguments

Value

List of upper and lower bounds for rho_uz

Solves for the variance of the error term u

Description

This function solves for the variance of u given r_TstarU and kappa. It handles 3 potential cases when the variance of u must be evaluated: 1. Across multiple simulations, but given the same r_TstarU and k 2. For multiple simulations, each with a value of r_TstarU and k 3. For one simulation across a grid of r_TstarU and k

Usage

get_s_u(r_TstarU, k, obs)

Arguments

Value

Vector of variances of u

Generates parameter estimates given user restrictions and data

Description

Generates parameter estimates given user restrictions and data

Usage

ivdoctr(
  y_name,
  T_name,
  z_name,
  data,
  example_name,
  controls = NULL,
  robust = FALSE,
  r_TstarU_restriction = c(-1, 1),
  k_restriction = c(1e-04, 1),
  n_draws = 5000,
  n_RF_draws = 1000,
  n_IS_draws = 1000,
  resample = FALSE
)

Arguments

Value

List with elements:

• ols: lm object of OLS estimation,

• iv: ivreg object of the IV estimation

• n: Number of observations

• b_OLS: OLS point estimate

• se_OLS: OLS standard errors

• b_IV: IV point estimate

• se_IV: IV standard errors

• k_lower: lower bound of kappa

• p_empty: fraction of parameter draws that yield an empty identified set

• p_valid: fraction of parameter draws compatible with a valid instrument

• r_uz_full_interval: 90% posterior credible interval for fully identified set of rho

• beta_full_interval: 90% posterior credible interval for fully identified set of beta

• r_uz_median: posterior median for partially identified rho

• r_uz_partial_interval: 90% posterior credible interval for partially identified set of rho under a conditionally uniform reference prior

• beta_median: posterior median for partially identified beta

• beta_partial_interval: 90% posterior credible interval for partially identified set of beta under a conditionally uniform reference prior

• a0: If treatment is binary, mis-classification probability of no-treatment case. NULL otherwise

• a1: If treatment is binary, mis-classification probability of treatment case. NULL otherwise

• psi_lower: lower bound for psi

• binary: logical indicating if treatment is binary

• k_restriction: User-specified bounds on kappa

• r_TstarU_restriction: User-specified bounds on r_TstarU

Examples

library(ivdoctr)
endog <- c(0, 0.9)
meas <- c(0.6, 1)

colonial_example1 <- ivdoctr(y_name = "logpgp95", T_name = "avexpr",
                            z_name = "logem4", data = colonial,
                            controls = NULL, robust = FALSE,
                            r_TstarU_restriction = endog,
                            k_restriction = meas,
                            example_name = "Colonial Origins")

Generates table of parameter estimates given user restrictions and data

Description

Generates table of parameter estimates given user restrictions and data

Usage

makeTable(..., output)

Arguments

Value

LaTeX code that generates output table with regression results

Examples

library(ivdoctr)
endog <- c(0, 0.9)
meas <- c(0.6, 1)

colonial_example1 <- ivdoctr(y_name = "logpgp95", T_name = "avexpr",
                            z_name = "logem4", data = colonial,
                            controls = NULL, robust = FALSE,
                            r_TstarU_restriction = endog,
                            k_restriction = meas,
                            example_name = "Colonial Origins")
makeTable(colonial_example1, output = file.path(tempdir(), "colonial.tex"))

Takes the OLS and IV estimates and converts it to a row of the LaTeX table

Description

Takes the OLS and IV estimates and converts it to a row of the LaTeX table

Usage

make_full_row(stats, example_name)

Arguments

Value

LaTeX code passed to makeTable()

Makes LaTeX code to make a row of a table and shift by some amount of columns if necessary

Description

Makes LaTeX code to make a row of a table and shift by some amount of columns if necessary

Usage

make_tex_row(char_vec, shift = 0)

Arguments

Value

LaTeX string of the whole row of the table

Generates a custom color palette given a vector of numbers

Description

Generates a custom color palette given a vector of numbers

Usage

map2color(x, pal, limits = NULL)

Arguments

Value

Hex values for colors

Rounds x to two decimal places

Description

Rounds x to two decimal places

Usage

myformat(x)

Arguments

Value

Number rounded to 2 decimal places

Plot ivdoctr Restrictions

Description

Plot ivdoctr Restrictions

Usage

plot_3d_beta(
  y_name,
  T_name,
  z_name,
  data,
  controls = NULL,
  r_TstarU_restriction = c(-1, 1),
  k_restriction = c(0, 1),
  n_grid = 30,
  n_colors = 500,
  fence = NULL,
  gray_k = NULL,
  gray_rTstarU = NULL,
  theta = 0,
  phi = 15
)

Arguments

Value

Interactive 3d plot which can be oriented and saved using rgl.snapshot()

Examples

library(ivdoctr)
endog <- matrix(c(0, 0.9), nrow = 1)
meas <- matrix(c(0.6, 1), nrow = 1)

plot_3d_beta(y_name = "logpgp95", T_name = "avexpr",
            z_name = "logem4", data = colonial,
            r_TstarU_restriction = endog,
            k_restriction = meas)

Construct vectors of points that outline a rectangle.

Description

Construct vectors of points that outline a rectangle.

Usage

rect_points(xleft, ybottom, xright, ytop, step_x, step_y)

Arguments

Value

List of x-coordinates and y-coordinates tracing the points around the rectangle

Simulate draws from the inverse Wishart distribution

Description

Simulate draws from the inverse Wishart distribution

Usage

rinvwish(n, v, S)

Arguments

Details

Employs the Bartlett Decomposition (Smith & Hocking 1972). Output exactly matches that of riwish from the MCMCpack package if the same random seed is used.

Value

A numeric array of matrices, each of which is one simulation draw.

Convert 3-d array to list of matrixes

Description

Convert 3-d array to list of matrixes

Usage

toList(myArray)

Arguments

Value

A list of numeric matrices.

Becker and Woessmann (2009) Dataset

Description

Data on Prussian counties in 1871 from Becker and Woessmann's (2009) paper "Was Weber Wrong? A Human Capital Theory of Protestant Economic History."

Usage

weber

Format

A data frame with 452 rows and 44 variables:

kreiskey1871

kreiskey1871

county1871

County name in 1871

rbkey

District key

lat_rad

Latitude (in rad)

lon_rad

Longitude (in rad)

kmwittenberg

Distance to Wittenberg (in km)

zupreussen

Year in which county was annexed by Prussia

hhsize

Average household size

gpop

Population growth from 1867-1871 in percentage points

f_prot

Percent Protestants

f_jew

Percent Jews

f_rw

Percent literate

f_miss

Percent missing education information

f_young

Percent below the age of 10

f_fem

Percent female

f_ortsgeb

Percent born in municipality

f_pruss

Percent of Prussian origin

f_blind

Percent blind

f_deaf

Percent deaf-mute

f_dumb

Percent insane

f_urban

Percent of county population in urban areas

lnpop

Natural logarithm of total population size

lnkmb

Natural logarithm of distance to Berlin (km)

poland

Dummy variable, =1 if county is Polish-speaking

latlon

Latitude * Longitude * 100

f_over3km

Percent of pupils farther than 3km from school

f_mine

Percent of labor force employed in mining

inctaxpc

Income tax revenue per capita in 1877

perc_secB

Percentage of labor force employed in manufacturing in 1882

perc_secC

Percentage of labor force employed in services in 1882

perc_secBnC

Percentage of labor force employed in manufacturing and services in 1882

lnyteacher

100 * Natural logarithm of male elementary school teachers in 1886

rhs

Dummy variable, =1 if Imperial of Hanseatic city in 1517

yteacher

Income of male elementary school teachers in 1886

pop

Total population size

kmb

Distance to Berlin (km)

uni1517

Dummy variable, =1 if University in 1517

reichsstadt

Dummy variable, =1 if Imperial city in 1517

hansestadt

Dummy variable, =1 if Hanseatic city in 1517

f_cath

Percentage of Catholics

sh_al_in_tot

Share of municipalities beginning with letter A to L

ncloisters1517_pkm2

Monasteries per square kilometer in 1517

school1517

Dummy variable, =1 if school in 1517

dnpop1500

City population in 1500

Source

https://www.ifo.de/en/iPEHD

References

https://www.ifo.de/en/iPEHD doi: 10.1162/qjec.2009.124.2.531