Citation

When using the CONSENSUS dataset please cite it using:

• Fernández-i-Marín, Knill, Steinebach (2021). Studying Policy Design Quality in Comparative Perspective. American Political Science Review. doi: 10.1017/S0003055421000186.

• Knill, Schulze & Tosun (2012). Regulatory Policy Outputs and Impacts. Exploring a Complex Relationship. Regulation & Governance, 5(4), 427-444. doi: 10.1111/j.1748-5991.2012.

When using the PolicyPortfolios package, please cite it using:

• Fernández-i-Marín (2020): Using PolicyPortfolios (http://xavier-fim.net/post/using_policyportfolios/).

Data release and and accompanying software

Both tools are available from the R statistical software and can be installed using:

install.packages("PolicyPortfolios")

The package contains the full set of social and environmental data in the form of a policy portfolio, as well as a set of functions developed to make its treatment easy for researchers.

To load the dataset you only need to load the package and the data:

library(PolicyPortfolios)
data(consensus)

## ℹ Loading PolicyPortfolios

## Package 'PolicyPortfolios' version 0.2.2
## Type 'citation("PolicyPortfolios")' for citing this R package
## or its datasets in publications.

Contents of the dataset

This loads the object consensus containing the full policy portfolios for the social and environmental sectors into R’s memory. A policy portfolio is a collection of simple assessments of the presence or absense of state intervention in a specific area (Target) using a concrete state capacity (Instrument).

The object is a tidy data frame (or tibble) with 6 columns, 5 identifying the unit (Sector, Country, Year, Instrument and Target) and 1 identifying whether such policy space is covered by public intervention or not.

str(consensus)

## tibble [509,220 × 6] (S3: tbl_df/tbl/data.frame)
##  $ Sector    : Factor w/ 2 levels "Environmental",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ Country   : Factor w/ 23 levels "Australia","Austria",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ Year      : int [1:509220] 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 ...
##  $ Instrument: Factor w/ 19 levels "Allowance","Bonus / Grant",..: 7 7 7 7 7 7 7 7 7 7 ...
##  $ Target    : Factor w/ 67 levels "Air quality standards for nitrogen oxides (Nox)",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ covered   : int [1:509220] 0 0 0 0 0 0 0 0 0 0 ...

consensus

The CONSENSUS dataset includes the following units:

• Sectors: Environmental, Social
• Countries: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, South Korea, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States
• Year range: 1976, 2005

The data release is completed with the list and meta-data of policy instruments (19 instruments, 13 in the environmental sector and 6 in the social sector):

data(consensus.instruments)
consensus.instruments

And also policy targets (67 targets, 49 in the environmental sector and 19 in the social sector):

data(consensus.targets)
consensus.targets

Convenient treatment of policy portfolios data

The set of functions in the package, coupled with R’s own functions makes the treatment of the date very easy.

Plotting a specific portfolio can be achieved by the function pp_plot specifying the Sector, Country and Year:

# Use the argument 'id' to pass a list of sectors, countries or years.
pp_plot(consensus, id = list(Sector = "Social", 
                             Country = "France",
                             Year = 1985))

Figure 1: Example of the use of the pp_plot() function, for the social portfolio in France in 1985.

As of version 0.2.2 (March 2021), the function pp_measures() allows to easily retrieve the following measures of the characteristics of the portfolios:

• Space: how many policy spaces does it contain (Instruments * Targets)
• Size: proportion of space covered by policy intervention (Adam, Knill, and Fernández-i-Marín 2017).
• n.Instruments: number of instruments covered at least by one target
• p.Instruments: proportion of instruments covered at least by one target
• n.Targets: number of targets covered at least by one instrument
• p.Targets: proportion of targets covered at least by one instrument
• Unique: number of unique instrument configurations
• Diversity (AID): Average Instrument Diversity (Fernández-i-Marín, Knill, and Steinebach 2021).
• Diversity (Gini-Simpson): (Simpson 1949) (Hill 1973)
• Diversity (Shannon): (Shannon 1948).
• Equality of instrument configurations.
• Equitability (Shannon): a variant of Shannon’s diversity.
• Instrument preponderance: on average, how many instruments per target.

For example, we can get a whole set of measures for the social portfolio in France in 1985.

# Use the argument 'id' to pass a list of sectors, countries or years
# for which to calculate the portfolio measures.
# It can also be achieved using pipes and filtering the data accordingly.
pp_measures(consensus, id = list(Sector = "Social", 
                                 Country = "France",
                                 Year = 1985))

Coupling with R’s working flow

The package works well with tidyverse packages, and its functions are easily coupled with the use of pipes (%>%) in R. For instance, plotting the portfolio size for two countries can be achieved with a combination of the package functions and the tidyverse tools:

library(ggplot2)
consensus %>%
  # Start with the whole dataset and limit its scope by sector and country
  filter(Sector == "Environmental") %>%
  filter(Country %in% c("Germany", "Italy")) %>%
  # Ensure that only the relevant factor levels are passed
  droplevels() %>%
  # calculate the measures for each Country and Year
  pp_measures() %>%
  # Only keep the size of the portfolio as a Measure to pass to ggplot()
  filter(Measure == "Size") %>%
  ggplot(aes(x = Year, y = value, color = Country)) +
  geom_line() +
  ylab("Portfolio size") +
  ggtitle("Environmental sector") +
  scale_color_manual(values = c("#E69F00", "#56B4E9"))

Figure 2: Environmental portfolio size for Germany and Italy over time.

Or summarizing the averages of all portfolio measures available by countries:

bind_rows(
  # Treat the environmental and social sectors separately
  # as they do not contain the same Instruments and Targets
  filter(consensus, Sector == "Environmental") %>%
    droplevels() %>%
    pp_measures(),
  filter(consensus, Sector == "Social") %>%
    droplevels() %>%
    pp_measures()) %>%
  # Once the measures have been calculated for each set of sectors
  # bind the rows together and calculate the average over countries
  group_by(Sector, Year, Measure.label) %>%
  summarize(Average = mean(value)) %>%
  # finally pass the data to ggplot()
  ggplot(aes(x = Year, y = Average, color = Sector)) +
  geom_line() +
  facet_wrap(~ Measure.label, scales = "free")

Figure 3: Temporal evolution of average measures of portfolio characteristics for the countries considered in the CONSENSUS dataset.

Further reading

You can learn more at:

• Package reference manual
• Vignette of the package
• Resources for policy portfolios