ConvergenceClub package

Subtitle

1 Basic Usage

data("countryGDP")
countryGDP

# Cluster Countries using GDP from year 2000 to year 2014, with 2014 as reference year
clubs <- findClubs(countryGDP, 
                   dataCol = 2:35, 
                   regions = 1, 
                   refCol = 35, 
                   time_trim = 1/3, 
                   cstar = 0, 
                   HACmethod = "AQSB")
summary(clubs)

Number of convergence clubs: 5
Number of divergent units: 3

             # of regions 
----------- ---------------
club1      | 84           
club2      | 18           
club3      | 26           
club4      | 14           
club5      | 7            
divergent  | 3            

# Merge clusters using Phillips and Sul (2009) method
mclubs <- mergeClubs(clubs, 
                     mergeMethod ='PS', 
                     mergeDivergent = FALSE)
summary(mclubs)

Number of convergence clubs: 4
Number of divergent units: 3

             # of regions  | merged clubs 
----------- --------------- --------------
club1      | 84            | clubs: 1     
club2      | 18            | clubs: 2     
club3      | 40            | clubs: 3, 4  
club4      | 7             | clubs: 5     
divergent  | 3             | clubs:       

# Merge clusters using von Lyncker and Thoennessen (2016) method
mclubs <- mergeClubs(clubs, 
                     mergeMethod = 'vLT', 
                     mergeDivergent = FALSE)
summary(mclubs)

Number of convergence clubs: 4
Number of divergent units: 3

             # of regions  | merged clubs 
----------- --------------- --------------
club1      | 84            | clubs: 1     
club2      | 18            | clubs: 2     
club3      | 26            | clubs: 3     
club4      | 21            | clubs: 4, 5  
divergent  | 3             | clubs:       

# Plot Transition Paths for all regions in each club and average Transition Path
# for all clubs
plot(mclubs)

# Plot Only average Transition Paths
plot(mclubs, clubs=NULL, legend=TRUE)

2 Using the World Income Database: ps2009.csv

ps2009 <- read_csv("https://raw.githubusercontent.com/ds777/sample-datasets/master/ps2009.csv")

Parsed with column specification:
cols(
  country = col_character(),
  year = col_integer(),
  pgdp = col_double()
)

ps2009

2.1 Reshaping from long to wide (controling names)

Useful reference for How to control new variables’ names after tidyr’s spread

ps2009 <- ps2009 %>% 
  mutate(year = paste('y', year, sep="")) %>% 
  spread(year, pgdp)
ps2009

2.2 Express variables in log terms

ps2009_log <- ps2009 %>% 
  mutate_if(is.numeric, funs("log" = log(.))) %>% 
  select(country, y1970_log:y2003_log)
ps2009_log

It would be ideal to have long-run trend of these variables, but for the moment, let us use them as they are.

2.3 Finding convergence clubs

Very important for the code to work: the data should be of type data.frame and not tbl_df

ps2009_log <- as.data.frame(ps2009_log)

clubs <- findClubs(ps2009_log,      # dataframe containing data 
                   dataCol = 2:35,  # integer vector with the column indices of the data
                   regions = 1,     # integer scalar indicating the column with codes of the regions
                   refCol = 35,     # integer scalar indicating the column to use for ordering data
                   time_trim = 1/3, # PS suggest to discard the first third of the period.
                   cstar = 0,       # the threshold value of the sieve criterion
                   HACmethod = "AQSB") # for the truncation of the Quadratic Spectral kernel
summary(clubs)

Number of convergence clubs: 5
Number of divergent units: 1

             # of regions 
----------- ---------------
club1      | 85           
club2      | 27           
club3      | 20           
club4      | 9            
club5      | 10           
divergent  | 1            

2.4 Merging clusters using Phillips and Sul (2009) method

mclubs_ps <- mergeClubs(clubs, 
                        mergeMethod ='PS', 
                        mergeDivergent = FALSE,
                        threshold = -1.65)
summary(mclubs_ps)

Number of convergence clubs: 4
Number of divergent units: 1

             # of regions  | merged clubs 
----------- --------------- --------------
club1      | 85            | clubs: 1     
club2      | 27            | clubs: 2     
club3      | 29            | clubs: 3, 4  
club4      | 10            | clubs: 5     
divergent  | 1             | clubs:       

2.5 Merging clusters using von Lyncker and Thoennessen (2016).

mclubs_vlt <- mergeClubs(clubs, 
                        mergeMethod ='vLT', 
                        mergeDivergent = FALSE,
                        threshold = -1.65)
summary(mclubs_vlt)

Number of convergence clubs: 4
Number of divergent units: 1

             # of regions  | merged clubs 
----------- --------------- --------------
club1      | 85            | clubs: 1     
club2      | 27            | clubs: 2     
club3      | 29            | clubs: 3, 4  
club4      | 10            | clubs: 5     
divergent  | 1             | clubs:       

2.6 Plot Transition Paths for all regions in each club and the average transition path for each club

For Phillips and Sul (2009)

plot(mclubs_ps)

For von Lyncker and Thoennessen (2016)

plot(mclubs_vlt)

2.7 Plot Only average Transition Paths

plot(mclubs_ps, clubs=NULL, legend=TRUE)

plot(mclubs_vlt, clubs=NULL, legend=TRUE)

3 References

• https://github.com/rhobis/ConvergenceClubs