nfcs.Rmd

# National Financial Capability Study (NFCS) {-}

[![License: GPL v3](https://img.shields.io/badge/License-GPLv3-blue.svg)](https://www.gnu.org/licenses/gpl-3.0) <a href="https://github.com/asdfree/nfcs/actions"><img src="https://github.com/asdfree/nfcs/actions/workflows/r.yml/badge.svg" alt="Github Actions Badge"></a> 

A study of financial knowledge and behavior, like making ends meet, planning ahead, managing assets.

* One state-by-state survey table with one row per sampled respondent, a separate investor survey.

* An online non-probability sample of U.S. adults (18+) calibrated to the American Community Survey.

* Released triennially since 2009.

* Funded by the [FINRA Investor Education Foundation](https://www.finrafoundation.org/) and conducted by [FGS Global](https://fgsglobal.com/).

---

## Recommended Reading {-}

Four Example Strengths & Limitations:

✔️ [Comprehensive assessment of financial literacy](https://gflec.org/education/questions-that-indicate-financial-literacy/)

✔️ [Questionnaire replicated by other studies](https://doi.org/10.1017/S1474747222000154)

❌ [Non-probability quota sampling from online panels](https://finrafoundation.org/sites/finrafoundation/files/NFCS-2009-StatebyState-Methodology.pdf)

❌ [Limited income and asset detail compared to CPS or SCF](https://finrafoundation.org/sites/finrafoundation/files/NFCS-2021-State-by-State-Questionnaire.pdf)

<br>

Three Example Findings:

1. [In 2018, 33% of Americans aged 51-61 were satisfied with their personal financial situations](https://www.nber.org/papers/w28236).

2. [The gender gap in financial literacy widened with age in 2021](https://dx.doi.org/10.2139/ssrn.4800263).

3. [Average scores on a test of five financial literacy questions declined between 2009 and 2021](https://finrafoundation.org/sites/finrafoundation/files/Why-Is-Measured-Financial-Literacy-Declining.pdf).

<br>

Two Methodology Documents:

> [2021 National Financial Capability Study: State-by-State Survey Methodology](https://finrafoundation.org/sites/finrafoundation/files/NFCS-2021-State-by-State-Methodology.pdf)

> [Financial Capability Insights: What the NFCS Reveals](https://www.finrafoundation.org/knowledge-we-gain-share/nfcs/other-research)

<br>

One Haiku:

```{r}
# lady madonna
# laid bank balance goose egg, loves
# gold unrequited
```

---

## Download, Import, Preparation {-}

Download and import the latest state-by-state microdata:
```{r eval = FALSE , results = "hide" }
library(haven)

zip_tf <- tempfile()

zip_url <- 
	'https://finrafoundation.org/sites/finrafoundation/files/2021-SxS-Data-and-Data-Info.zip'

download.file( zip_url , zip_tf , mode = 'wb' )

unzipped_files <- unzip( zip_tf , exdir = tempdir() )

stata_fn <- grep( "\\.dta$" , unzipped_files , value = TRUE )

nfcs_tbl <- read_dta( stata_fn )

nfcs_df <- data.frame( nfcs_tbl )

names( nfcs_df ) <- tolower( names( nfcs_df ) )
```

Add a column of all ones, add labels to state names, add labels to the rainy day fund question:
```{r eval = FALSE , results = "hide" }
nfcs_df[ , 'one' ] <- 1

nfcs_df[ , 'state_name' ] <-
	factor(
		nfcs_df[ , 'stateq' ] , 
		levels = 1:51 , 
		labels = sort( c( 'District of Columbia' , state.name ) ) 
	)

nfcs_df[ , 'rainy_day_fund' ] <-
	factor(
		nfcs_df[ , 'j5' ] ,
		levels = c( 1 , 2 , 98 , 99 ) ,
		labels = c( 'Yes' , 'No' , "Don't Know" , "Prefer not to say" )
	)

```

### Save Locally \ {-}

Save the object at any point:

```{r eval = FALSE , results = "hide" }
# nfcs_fn <- file.path( path.expand( "~" ) , "NFCS" , "this_file.rds" )
# saveRDS( nfcs_df , file = nfcs_fn , compress = FALSE )
```

Load the same object:

```{r eval = FALSE , results = "hide" }
# nfcs_df <- readRDS( nfcs_fn )
```

### Survey Design Definition {-}
Construct a complex sample survey design:

```{r eval = FALSE , results = "hide" }

library(survey)

nfcs_design <- svydesign( ~ 1 , data = nfcs_df , weights = ~ wgt_n2 )

divison_design <- svydesign( ~ 1 , data = nfcs_df , weights = ~ wgt_d2 )

state_design <- svydesign( ~ 1 , data = nfcs_df , weights = ~ wgt_s3 )
```

### Variable Recoding {-}

Add new columns to the data set:
```{r eval = FALSE , results = "hide" }
nfcs_design <- 
	update( 
		nfcs_design ,
		
		satisfaction_w_finances =
			ifelse( j1 > 10 , NA , j1 ) ,
			
		risk_taking =
			ifelse( j2 > 10 , NA , j2 ) ,
		
		difficult_to_pay_bills =
			factor(
				j4 ,
				levels = c( 1 , 2 , 3 , 98 , 99 ) ,
				labels = 
					c( 
						'Very difficult' , 
						'Somewhat difficult' , 
						'Not at all difficult' , 
						"Don't know" , 
						'Prefer not to say' 
					)
			) ,
				
		spending_vs_income =
			factor(
				j3 ,
				levels = c( 1 , 2 , 3 , 98 , 99 ) ,
				labels = 
					c( 
						'Spending less than income' , 
						'Spending more than income' , 
						'Spending about equal to income' , 
						"Don't know" , 
						'Prefer not to say' 
					)
			) ,
		
		unpaid_medical_bills =
			ifelse( g20 > 2 , NA , as.numeric( g20 == 1 ) )
	)
```

---

## Analysis Examples with the `survey` library \ {-}

### Unweighted Counts {-}

Count the unweighted number of records in the survey sample, overall and by groups:
```{r eval = FALSE , results = "hide" }
sum( weights( nfcs_design , "sampling" ) != 0 )

svyby( ~ one , ~ spending_vs_income , nfcs_design , unwtd.count )
```

### Weighted Counts {-}
Count the weighted size of the generalizable population, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ one , nfcs_design )

svyby( ~ one , ~ spending_vs_income , nfcs_design , svytotal )
```

### Descriptive Statistics {-}

Calculate the mean (average) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svymean( ~ satisfaction_w_finances , nfcs_design , na.rm = TRUE )

svyby( ~ satisfaction_w_finances , ~ spending_vs_income , nfcs_design , svymean , na.rm = TRUE )
```

Calculate the distribution of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svymean( ~ difficult_to_pay_bills , nfcs_design )

svyby( ~ difficult_to_pay_bills , ~ spending_vs_income , nfcs_design , svymean )
```

Calculate the sum of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ satisfaction_w_finances , nfcs_design , na.rm = TRUE )

svyby( ~ satisfaction_w_finances , ~ spending_vs_income , nfcs_design , svytotal , na.rm = TRUE )
```

Calculate the weighted sum of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svytotal( ~ difficult_to_pay_bills , nfcs_design )

svyby( ~ difficult_to_pay_bills , ~ spending_vs_income , nfcs_design , svytotal )
```

Calculate the median (50th percentile) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
svyquantile( ~ satisfaction_w_finances , nfcs_design , 0.5 , na.rm = TRUE )

svyby( 
	~ satisfaction_w_finances , 
	~ spending_vs_income , 
	nfcs_design , 
	svyquantile , 
	0.5 ,
	ci = TRUE , na.rm = TRUE
)
```

Estimate a ratio:
```{r eval = FALSE , results = "hide" }
svyratio( 
	numerator = ~ satisfaction_w_finances , 
	denominator = ~ risk_taking , 
	nfcs_design ,
	na.rm = TRUE
)
```

### Subsetting {-}

Restrict the survey design to persons receiving pandemic-related stimulus payment:
```{r eval = FALSE , results = "hide" }
sub_nfcs_design <- subset( nfcs_design , j50 == 1 )
```
Calculate the mean (average) of this subset:
```{r eval = FALSE , results = "hide" }
svymean( ~ satisfaction_w_finances , sub_nfcs_design , na.rm = TRUE )
```

### Measures of Uncertainty {-}

Extract the coefficient, standard error, confidence interval, and coefficient of variation from any descriptive statistics function result, overall and by groups:
```{r eval = FALSE , results = "hide" }
this_result <- svymean( ~ satisfaction_w_finances , nfcs_design , na.rm = TRUE )

coef( this_result )
SE( this_result )
confint( this_result )
cv( this_result )

grouped_result <-
	svyby( 
		~ satisfaction_w_finances , 
		~ spending_vs_income , 
		nfcs_design , 
		svymean ,
		na.rm = TRUE 
	)
	
coef( grouped_result )
SE( grouped_result )
confint( grouped_result )
cv( grouped_result )
```

Calculate the degrees of freedom of any survey design object:
```{r eval = FALSE , results = "hide" }
degf( nfcs_design )
```

Calculate the complex sample survey-adjusted variance of any statistic:
```{r eval = FALSE , results = "hide" }
svyvar( ~ satisfaction_w_finances , nfcs_design , na.rm = TRUE )
```

Include the complex sample design effect in the result for a specific statistic:
```{r eval = FALSE , results = "hide" }
# SRS without replacement
svymean( ~ satisfaction_w_finances , nfcs_design , na.rm = TRUE , deff = TRUE )

# SRS with replacement
svymean( ~ satisfaction_w_finances , nfcs_design , na.rm = TRUE , deff = "replace" )
```

Compute confidence intervals for proportions using methods that may be more accurate near 0 and 1. See `?svyciprop` for alternatives:
```{r eval = FALSE , results = "hide" }
svyciprop( ~ unpaid_medical_bills , nfcs_design ,
	method = "likelihood" , na.rm = TRUE )
```

### Regression Models and Tests of Association {-}

Perform a design-based t-test:
```{r eval = FALSE , results = "hide" }
svyttest( satisfaction_w_finances ~ unpaid_medical_bills , nfcs_design )
```

Perform a chi-squared test of association for survey data:
```{r eval = FALSE , results = "hide" }
svychisq( 
	~ unpaid_medical_bills + difficult_to_pay_bills , 
	nfcs_design 
)
```

Perform a survey-weighted generalized linear model:
```{r eval = FALSE , results = "hide" }
glm_result <- 
	svyglm( 
		satisfaction_w_finances ~ unpaid_medical_bills + difficult_to_pay_bills , 
		nfcs_design 
	)

summary( glm_result )
```

---

## Replication Example {-}

This example matches the unweighted count shown on [PDF page 4](https://finrafoundation.org/sites/finrafoundation/files/NFCS-Report-Fifth-Edition-July-2022.pdf#page=4):

```{r eval = FALSE , results = "hide" }
stopifnot( nrow( nfcs_df ) == 27118 )
```

This example matches the [PDF page 7](https://finrafoundation.org/sites/finrafoundation/files/NFCS-Report-Fifth-Edition-July-2022.pdf#page=7) estimate that _53% have three months of rainy day funds_:
```{r eval = FALSE , results = "hide" }
national_rainy_day <- svymean( ~ rainy_day_fund , nfcs_design )
stopifnot( round( coef( national_rainy_day )[ 'rainy_day_fundYes' ] , 2 ) == 0.53 )
```

This example matches counts and rainy day estimates from [The Geography of Financial Capability](https://cdn.finra.org/nfcs/2021/geography.html):
```{r eval = FALSE , results = "hide" }
state_counts <-
	svyby(
		~ one ,
		~ state_name ,
		state_design ,
		unwtd.count
	)
	
stopifnot( state_counts[ 'California' , 'counts' ] == 1252 )
stopifnot( state_counts[ 'Missouri' , 'counts' ] == 501 )
stopifnot( state_counts[ 'Oregon' , 'counts' ] == 1261 )

state_rainy_day <-
	svyby(
		~ rainy_day_fund ,
		~ state_name ,
		state_design ,
		svymean
	)
	
stopifnot( round( state_rainy_day[ 'California' , 'rainy_day_fundYes' ] , 2 ) == 0.57 )
stopifnot( round( state_rainy_day[ 'Missouri' , 'rainy_day_fundYes' ] , 2 ) == 0.51 )
stopifnot( round( state_rainy_day[ 'Oregon' , 'rainy_day_fundYes' ] , 2 ) == 0.52 )
```

---

## Analysis Examples with `srvyr` \ {-}

The R `srvyr` library calculates summary statistics from survey data, such as the mean, total or quantile using [dplyr](https://github.com/tidyverse/dplyr/)-like syntax. [srvyr](https://github.com/gergness/srvyr) allows for the use of many verbs, such as `summarize`, `group_by`, and `mutate`, the convenience of pipe-able functions, the `tidyverse` style of non-standard evaluation and more consistent return types than the `survey` package. [This vignette](https://cran.r-project.org/web/packages/srvyr/vignettes/srvyr-vs-survey.html) details the available features. As a starting point for NFCS users, this code replicates previously-presented examples:

```{r eval = FALSE , results = "hide" }
library(srvyr)
nfcs_srvyr_design <- as_survey( nfcs_design )
```
Calculate the mean (average) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
nfcs_srvyr_design %>%
	summarize( mean = survey_mean( satisfaction_w_finances , na.rm = TRUE ) )

nfcs_srvyr_design %>%
	group_by( spending_vs_income ) %>%
	summarize( mean = survey_mean( satisfaction_w_finances , na.rm = TRUE ) )
```