Index of Economic Freedom Comparisons

Introduction

In this document I contrast country scores in the Heritage Foundation Index of Economic Freedom, focusing on the U.S. versus the Nordic countries Denmark, Finland, Norway, and Sweden, with New Zealand included as an example of a top-ranking country. (I don’t include Iceland because of its very small population.)

This document includes the code to produce the graph heading my post on social democracy published as part of my “Seven Answers” series of blog posts.

Setup and data preparation

Libraries

I use the tidyverse set of functions for general data manipulation and the knitr package to create a formatted table.

library("tidyverse")
library("knitr")

Data sources

I use the following CSV file in this analysis:

ief-scores-1995-2018.csv contains the full list of scores for all countries for the years 1995 through 2018.

For more information on how I created this file see the “References” section.

Each row of the file contains the following 15 variables (“chr” indicates a character string, “int” an integer value, and “num” a numeric value with decimal point):

name (chr). The country name (e.g., “United States”) associated with the scores.
index_year (int). The year in which the index was published (e.g, 2018). This is usually the year after the year of data from which the component scores are calculated.
overall_score (num). The average of the 12 component score variables for a given index year and country.
property_rights (num). This and the following component scores run from 0 to 100, and are rounded to 1 digit past the decimal point (thus 3 significant digits).
government_integrity (num).
judicial_effectiveness (num).
tax_burden (num).
government_spending (num).
fiscal_health (num).
business_freedom (num).
labor_freedom (num).
monetary_freedom (num).
trade_freedom (num).
investment_freedom (num).
financial_freedom (num).

Reading in and formatting the data

I begin by reading the CSV file ief-scores-1995-2018.csv into the table ief. The file does not need any further cleaning or revision before being used.

ief <- read_csv("ief-scores-1995-2018.csv")

Analysis

I do one main plot and three other analyses just for fun:

A comparison of overall scores between the U.S., Nordic countries, and New Zealand over the entire history of the Index of Economic Freedom.
A comparison of component scores between the U.S., Nordic countries, and New Zealand for the 2018 index.
An alternate ranking of countries for 2018, leaving out the spending and taxation components (government_spending and tax_burden).
A companion graph to the first one, showing a comparison of overall scores over time if the spending and taxation components are removed.

Comparison of overall scores for 1995-2018

For the main graph I want only the overall scores, so I create a second table ief_overall with that data for all years and countries.

ief_overall <- ief %>%
  select(name, index_year, overall_score)

I plot the data by year, showing a separate line for each country of interest. I include all years from 1995 on. (The IEF methodology changed over the years, so scores from earlier years aren’t strictly comparable with scores from later years. However, I’m just looking for overall trends.)

To color the graph lines I use a color palette designed to be more visible for people with color blindness. The palette is set so as to match up with the (alphabetical) order of countries in the legend, so that I can highlight the U.S. score in black and the New Zealand score in gray.

palette <- c("#E69F00", "#56B4E9", "#999999",
             "#009E73", "#CC79A7", "#000000")

countries <- c("Denmark", "Finland", "New Zealand",
               "Norway", "Sweden", "United States")

ief_overall %>%
  filter(index_year >= 1995) %>%
  filter(name %in% countries) %>%
  ggplot(mapping=aes(x=index_year, y=overall_score,
                     group=name, color=name)) +
    geom_line(size=0.8) +
    scale_color_manual(values=palette) +
    coord_cartesian(ylim=c(60, 90)) +
    scale_x_continuous(breaks=seq(1995,2020,5)) +
    labs(x="Year", y="Economic Freedom Score", color="Country") +
    theme_bw()

Comparison of component scores

To investigate the extent to which the various components of the overall score differ from country to country, I create a table ief_subscores containing the scores for the various components that go into the overall scores for 2018, using the following procedure:

Filter the data to use only the rows for 2018.
Filter again to use only the rows for the countries of interest.
Get rid of the index_year and overall_score columns, which we don’t need.
Gather the scores for each component freedom together in a single column score with a new column component to hold the name of the particular component of the overall score.
Divide the scores by 12 to reflect the fact that they are averaged together to create the overall score.

ief_subscores <- ief %>%
  filter(index_year == 2018) %>%
  filter(name %in% countries) %>%
  select(-index_year, -overall_score) %>%
  gather(component, score, -name) %>%
  mutate(score=score/12)

Finally, I create a stacked bar chart showing how the component scores add together to produce the overall scores for each country.

I use an alternate palette for the component fills because the default palette makes it difficult to see how much each component affects the overall result. (The colors repeat because there are more components than colors in the palette, but since the position of the components in the stacks matches the position of the corresponding components names in the legend, it’s still reasonably clear which component is which.)

component_palette <- c("#999999", "#E69F00", "#56B4E9", "#009E73",
                       "#F0E442", "#0072B2", "#D55E00", "#CC79A7",
                       "#999999", "#E69F00", "#56B4E9", "#009E73")

ief_subscores %>%
  ggplot(mapping=aes(x=name, y=score, fill=component)) +
    geom_bar(stat='identity') +
    scale_fill_manual(values = component_palette) +
    labs(x="Country",
         y="Components of Economic Freedom Score",
         fill="Component") +
  theme_bw()

Recall that the thinner the bar for a component the lower the component score and the more the country is being penalized with respect to that component.

Alternative analysis ignoring government spending and taxation

By looking at the component scores I see that some Nordic countries are penalized for their relatively high taxes and levels of government spending. How do the rankings change if we remove these components?

To find out, I first create a table of countries ranked in order of the values of overall_score in 2018. (This corresponds to the official rankings as published on the Heritage Foundation site.) I do this as follows:

Filter the data for scores for 2018.
Keep only the name and overall_score columns.
Sort the rows in descending order of overall_score (highest to lowest).
Add a new variable giving each country’s ranking (its row number after sorting).

rankings_2018 <- ief %>%
  filter(index_year == 2018) %>%
  select(name, overall_score) %>%
  arrange(desc(overall_score)) %>%
  mutate(ranking = row_number())

Next I compute an alternative overall score using only ten components, with government_spending and tax_burden removed, and then create an alternative ranking from that. I do this as follows:

Filter the data for scores for 2018.
Remove the government_spending and tax_burden columns.
Compute a new overall_score variable as the average of the last 10 columns, rounded to 1 digit past the decimal place. (I ignore missing values so that they don’t affect the average.)
Sort the rows in descending order of overall_score (highest to lowest).
Keep only the name and overall_score columns.
Sort the rows in descending order of overall_score (highest to lowest).
Add a new variable giving each country’s ranking (its row number after sorting).

alt_rankings_2018 <- ief %>%
  filter(index_year == 2018) %>%
  select(-government_spending, -tax_burden) %>%
  mutate(overall_score = round(rowMeans(.[4:13], na.rm=TRUE), 1)) %>%
  select(name, overall_score) %>%
  arrange(desc(overall_score)) %>%
  mutate(ranking = row_number())

Then for convenience of comparison I create a table that has both sets of rankings side by side. I do this as follows:

Take the alternative table and rename the name and overall_score columns, because I want to keep them separate from the columns of the same name in the original rankings.
Join the two tables, with ranking as the common variable.
Sort the joined table based on the ranking value.
Select the columns in the order I want to see them.

comparison_rankings_2018 <- alt_rankings_2018 %>%
  rename(alt_name = name, alt_overall_score = overall_score) %>%
  inner_join(rankings_2018) %>%
  arrange(ranking) %>%
  mutate(alt_ranking = ranking) %>%
  select(ranking, name, overall_score,
         alt_ranking, alt_name, alt_overall_score)

Finally, I display a table of the two sets of rankings compared; for brevity I show only the first 30 countries, which in the original rankings includes all the Nordic countries.

comparison_rankings_2018 %>%
  filter(ranking <= 30) %>%
  rename(Rank=ranking,
         Country=name,
         Score=overall_score,
         "Rank (Alternative Ranking)"=alt_ranking,
         "Country (Alternative Ranking)"=alt_name,
         "Score (Alternative Ranking)"=alt_overall_score) %>%
  kable()

Rank	Country	Score	Rank (Alternative Ranking)	Country (Alternative Ranking)	Score (Alternative Ranking)
1	Hong Kong	90.2	1	Hong Kong	90.0
2	Singapore	88.8	2	New Zealand	89.1
3	New Zealand	84.2	3	Singapore	88.4
4	Switzerland	81.7	4	Denmark	86.8
5	Australia	80.9	5	Liechtenstein	85.0
6	Ireland	80.4	6	Sweden	84.8
7	Estonia	78.8	7	Australia	84.6
8	United Kingdom	78.0	8	Switzerland	84.5
9	Canada	77.7	9	United Kingdom	82.6
10	United Arab Emirates	77.6	10	Netherlands	82.3
11	Iceland	77.0	11	Finland	82.1
12	Denmark	76.6	12	Ireland	81.9
13	Taiwan	76.6	13	Estonia	81.2
14	Luxembourg	76.4	14	Iceland	80.7
15	Sweden	76.3	15	Norway	80.6
16	Georgia	76.2	16	Canada	80.3
17	Netherlands	76.2	17	Luxembourg	80.3
18	United States	75.7	18	Austria	79.2
19	Lithuania	75.3	19	Germany	78.8
20	Chile	75.2	20	United States	78.7
21	Mauritius	75.1	21	United Arab Emirates	76.2
22	Malaysia	74.5	22	Czech Republic	76.0
23	Norway	74.3	23	Belgium	75.4
24	Czech Republic	74.2	24	Georgia	75.4
25	Germany	74.2	25	Israel	75.4
26	Finland	74.1	26	Lithuania	75.3
27	South Korea	73.8	27	Taiwan	75.3
28	Latvia	73.6	28	Japan	74.6
29	Qatar	72.6	29	South Korea	74.4
30	Japan	72.3	30	Chile	74.3

Note the dramatic improvements in rankings for the Nordic countries:

Denmark has jumped from position 12 to position 4.
Sweden has jumped from position 15 to position 6.
Finland has jumped from position 26 to position 11.
Norway has jumped from position 23 to position 15.

However,the United States remains almost unchanged, dropping from position 18 to position 20.

Comparison of alternate scores for 1995-2018 (without spending and taxation)

What would the comparison graph above look like if we ignored the spending and taxation components of the overall score. Here I repeat the original graph using an alternative score. I do this as follows:

Start with the original full data set.
Filter for rows for the countries we’re interested in.
Remove the government_spending and tax_burden variables from the data set.
Compute an alternative score alt_overall_score as we did above, taking the average of the remaining 10 variables. (If there are any missing values we ignore them and just take the average of the variables that do have values.)
Select the variables of interest: name, index_year, and alt_overall_score.
Plot the variables as we did in the first graph.

The countries and palette variables are carried over from the original plot.

ief %>%
  filter(name %in% countries) %>%
  select(-government_spending, -tax_burden) %>%
  mutate(alt_overall_score = round(rowMeans(.[4:13], na.rm=TRUE), 1)) %>%
  select(name, index_year, alt_overall_score) %>%
  ggplot(mapping=aes(x=index_year, y=alt_overall_score,
                     group=name, color=name)) +
    geom_line(size=0.8) +
    scale_color_manual(values=palette) +
    coord_cartesian(ylim=c(60, 95)) +
    scale_x_continuous(breaks=seq(1995,2020,5)) +
    labs(x="Year", y="Alternate Economic Freedom Score", color="Country") +
    theme_bw()

In this alternative perspective Denmark was significantly more economically free than New Zealand for most of the last 10-15 years, losing ground only recently. (In fact, I had to increase the upper limit of the graph from 90 to 95 to avoid cutting off Denmark’s score.) Sweden has been comparable to or better than the United States in economic freedom since the early 2000s, and both Finland and Norway have caught up to and surpassed the U.S. in recent years.

Appendix

References

Unfortunately the Heritage Foundation does not provide an easy way to download all scores at once for all years. I worked around this using the following procedure to create the CSV file ief-scores-1995-2018.csv:

Go to the Data Explorer page and select the tab “All Index Data”.
For option 1, “Choose individual countries and/or regions”, select “All” in the region list.
For option 2, “Select year(s)”, select all years. (You must do this by shift-clicking the individual year numbers; you cannot select the “All” option.)
Click the button “View the data”.
On the resulting screen select all data displayed, all rows and all columns, and copy them to the clipboard.
Paste the data from the clipboard into an empty Excel spreadsheet.
Save the spreadsheet in CSV format.
Convert the file to use Unix line endings.
Replace spaces with underscores in the header variable names.
Change all occurrences of “N/A” to “NA”.

Environment

I used the following R environment in doing the analysis above:

sessionInfo()

## R version 3.4.4 (2018-03-15)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 16.04.4 LTS
## 
## Matrix products: default
## BLAS: /usr/lib/libblas/libblas.so.3.6.0
## LAPACK: /usr/lib/lapack/liblapack.so.3.6.0
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
## [1] bindrcpp_0.2    knitr_1.17      dplyr_0.7.2     purrr_0.2.3    
## [5] readr_1.1.1     tidyr_0.7.0     tibble_1.3.3    ggplot2_2.2.1  
## [9] tidyverse_1.1.1
## 
## loaded via a namespace (and not attached):
##  [1] Rcpp_0.12.12     highr_0.6        cellranger_1.1.0 compiler_3.4.4  
##  [5] plyr_1.8.4       bindr_0.1        forcats_0.2.0    tools_3.4.4     
##  [9] digest_0.6.12    lubridate_1.6.0  jsonlite_1.5     evaluate_0.10.1 
## [13] nlme_3.1-131.1   gtable_0.2.0     lattice_0.20-35  pkgconfig_2.0.1 
## [17] rlang_0.1.2      psych_1.7.5      yaml_2.1.14      parallel_3.4.4  
## [21] haven_1.1.0      xml2_1.1.1       httr_1.3.0       stringr_1.2.0   
## [25] hms_0.3          tidyselect_0.1.1 rprojroot_1.2    grid_3.4.4      
## [29] glue_1.1.1       R6_2.2.2         readxl_1.0.0     foreign_0.8-69  
## [33] rmarkdown_1.6    modelr_0.1.1     reshape2_1.4.2   magrittr_1.5    
## [37] backports_1.1.0  scales_0.4.1     htmltools_0.3.6  rvest_0.3.2     
## [41] assertthat_0.2.0 mnormt_1.5-5     colorspace_1.3-2 labeling_0.3    
## [45] stringi_1.1.5    lazyeval_0.2.0   munsell_0.4.3    broom_0.4.2

You can find the source code for this analysis and others at my Seven Answers public Gitlab repository. This document and its source code are available for unrestricted use, distribution and modification under the terms of the Creative Commons CC0 1.0 Universal (CC0 1.0) Public Domain Dedication. Stated more simply, you’re free to do whatever you’d like with it.