setwd("~/Desktop/Program Eval")
library(readr)
library(rdrobust)
census <- read_csv("census.csv")## Rows: 3139 Columns: 22
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## dbl (22): oldcode, hsspend_per_cap68_z1, hsspend_per_cap68_z2, pop60, povrat...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.mortality <- read_csv("mortality.csv")## Rows: 2810 Columns: 11
## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## dbl (11): povrate60, mort_age59_related_postHS, census1960_pop, census1960_p...
##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.Part 2 - 2/20/26 Question 3: Use the data census.csv and the rdplot command to check whether there is a discontinuity in Head Start spending per 4-year old in 1968 at the poverty rate cutoff of 59.1984%. Note that the variable povrate60 gives the poverty rate for the county and hsspend per kid 68 gives the Head Start spending per 4-year old child in 1968. Is there a jump in Head Start funding right at the cutoff for qualifying for grant writing help? If so, approximately how big is the jump? You do not need to comment on statistical significance.
rdplot(census$hsspend_per_kid_68, census$povrate60, c = 59.1984,
x.label = "County Poverty Rate (1960)",
y.label = "Head Start Spending per Kid (1968)")
There is a jump of about $150-200.
Question 4: Is there a discontinuity in overall social services spending per capita in 1972 (socspend per cap72) at the poverty rate cutoff of 59.1984%? Use the rdplot command to check. You do not need to comment on statistical significance. In what sense is this a “placebo” check?
rdplot(census$socspend_per_cap72, census$povrate60, c = 59.1984,
x.label = "County Poverty Rate (1960)",
y.label = "Other Social Services Spendin (1972)")
There is no jump. This is a placebo check to make sure there aren’t
other county variables impacting this except the head start grant
support. It is smooth through the cutoff.
Question 5: Use the data mortality.csv and the rdplot command to generate a plot showing mort age59 related postHS around the 59.1984% cutoff. Is there evidence in the plot you generated that Head Start participation reduces childhood mortality from Head Start related causes? You do not need to comment on statistical significance.
rdplot(mortality$mort_age59_related_postHS, mortality$povrate60, c = 59.1984,
x.label = "County Poverty Rate (1960)",
y.label = "Related Age 5-9 Mortality (per 100,000)")
There is a decrease in mortality among those who received the Head Start funding.
Question 6:Use the rdrobust command to estimate a regression that estimates how much Head Start spending per 4-year old in 1968 jumps at the discontinuity of 59.1984%. You will need to use the Census dataset for this question. Interpret the coefficient and comment on its statistical significance.
firststage <- rdrobust(census$hsspend_per_kid_68, census$povrate60, c = 59.1984)
summary(firststage)## Sharp RD estimates using local polynomial regression.
##
## Number of Obs. 3132
## BW type mserd
## Kernel Triangular
## VCE method NN
##
## Number of Obs. 2832 300
## Eff. Number of Obs. 274 201
## Order est. (p) 1 1
## Order bias (q) 2 2
## BW est. (h) 7.743 7.743
## BW bias (b) 13.342 13.342
## rho (h/b) 0.580 0.580
## Unique Obs. 2830 300
##
## =====================================================================
## Point Robust Inference
## Estimate z P>|z| [ 95% C.I. ]
## ---------------------------------------------------------------------
## RD Effect 135.231 1.410 0.159 [-56.368 , 345.048]
## =====================================================================There are 3132 counties in our dataset. 2832 of the counties are below the cuttoff. 300 are above the cutoff. According to bandwith, we are using 274 below and 201 observations above. At this cutoff, the head start funding was $135 higher but this is not statistically significant.
Question 7: Use the rdrobust command to estimate a regression that estimates whether the 5-9 year old mortality rate changes at the discontinuity of 59.1984%. You will need to use the mortality dataset for this question. Interpret the coefficient and comment on its statistical significance
reducedform <- rdrobust(mortality$mort_age59_related_postHS, mortality$povrate60, c = 59.1984)
summary(reducedform)## Sharp RD estimates using local polynomial regression.
##
## Number of Obs. 2783
## BW type mserd
## Kernel Triangular
## VCE method NN
##
## Number of Obs. 2489 294
## Eff. Number of Obs. 234 180
## Order est. (p) 1 1
## Order bias (q) 2 2
## BW est. (h) 6.810 6.810
## BW bias (b) 10.725 10.725
## rho (h/b) 0.635 0.635
## Unique Obs. 2489 294
##
## =====================================================================
## Point Robust Inference
## Estimate z P>|z| [ 95% C.I. ]
## ---------------------------------------------------------------------
## RD Effect -2.409 -2.032 0.042 [-5.462 , -0.099]
## =====================================================================Getting the grant writing support to get head start prevent 2.4 deaths out of 100,000 that would have otherwise happened. This is significant at the % level.