Working with Crotian Symmetric Input-Output Tables

Croatia

To show how to work with other IO tables and other data, we include the latest SIOT and employment statistics from Croatia with the package. Croatia joined the EU after 2010 and the country’s SIOTs are following the Eurostat guidelines but are not available on Eurostat yet. The national Symmetric Input-Output Tables (SIOTs) are created every five years, with a very long statistical work. For most EU countries the latest SIOT available is based on the 2010 structure of the underlying national economy. Because Croatia was not a member state at the time, it uses a temporary SIOT that is identical in most aspects to the Eurostat standard SIOTs.

There are some technical differences that can be obtains from DZS. The most important practical differences are:

• The SIOT is not available on Eurostat, and no reproducible function exists to get it. To overcome this shortfall, the SIOT is included in the package as packaged data with code how it was imported from the official Excel file.

• The data is not given in million but thousand national currency units (i.e. kunas). For some analytical work this has no practical differences (for example, the calculation of some multipliers.) But when you use currency units in your equation, recognize the difference.

The data is provided in the same long form that is used by Eurostat. The data file contains all the three parts of the SIOT and aggregation rows.

The strategy to work with SIOTs in Excel format is the following: - Read in the Excel table with any suitable R excel reader. - Use t_rows2 columns to describe the rows of the table, use t_cols2 to describe the columns of the table. - You can additionally add longer labels called t_rows2_lab and t_cols2_lab. - Bring the data to a tidy, long-form, for example with plyr or dplyr functions. - Filter out the part of the SIOT you need for your matrix equations with the iotables pre-processing functions.

You should aim at something similar to croatia_2010_1700, the table 1700 for the year 2010.

library (iotables); library (dplyr); require (knitr)
hr_io_1800 <- iotable_get ( source = "croatia_2010_1800", geo = "HR",
                         year = 2010, unit = "T_NAC", 
                         labelling  = "short")


knitr::kable(head(hr_io_1800, 3))

Preapring the matrixes for analytical work. By default, the totalling columns are removed from the matrixes. You can leave them for presentation purposes, but for further analytical work they need to be removed.

require(iotables); require (knitr)
hr_use_1800 <- use_table_get ( source = "croatia_2010_1800", geo = "HR",
                               year = 2010, unit = "T_NAC", 
                               labelling = "iotables", 
                               keep_total = FALSE)
output_vector_hr <- output_get(source = "croatia_2010_1800", geo = "HR",
                               year = 2010, unit = "T_NAC", 
                               labelling = "iotables", 
                               keep_total = FALSE)
knitr::kable(head(hr_use_1800, 3))

As a first analytical step we create the input coefficient matrix which is the input of the Leontieff-matrix and the Leontieff-inverse. In this case the original table created by DZS contains an empty column / row pair for “extraterriorial_organizations” which causes an error in the inversion of the Leontieff-matrix. We carefully remove both the column and the row to maintain the symmetry of the matrix and create a meaningful Leontieff-inverse. Keep in mind that for any further additional data you must remove “extraterriorial_organizations” organizations.

require(iotables); require (knitr)
hr_coeff <- input_coefficient_matrix_create( 
  input_flow = hr_use_1800,
  output = output_vector_hr, 
  digits = 6)

#In case we used 'short' type labels:
#hr_coeff <- hr_coeff %>%
#  select ( -U ) %>%
#  filter ( t_rows2 != "CPA_U")

#In case we used 'iotables' type labels:
hr_coeff <- hr_coeff %>%
  select ( -extraterriorial_organizations ) %>%
  filter ( iotables_row != "extraterriorial_organizations")

L_hr <- leontieff_matrix_create( technology_coefficients_matrix =
                                hr_coeff )
I_hr <- leontieff_inverse_create(L_hr)
knitr::kable(head(hr_coeff, 3))

You cannot rely on the input_indicator_create function because the official Croatian input-output tables do not include employment data. Instead we created the following table from the official Croatian employment data. The creation of such auxilliary tables may be needed in other analysis, and it is quiet straightforward but requires judgement and manual work.

You need to aggregate the data, in this case the employment data, exactly into the industry / product groups of the IO table. In this lucky case, the employment data has a higher resolution, and we need to add various industry / product groups togeher. For example, the CPA_F industry group (construction) is given for three subindustries in the employment statistics of Croatia that we had to simply add together.

If you have less detailed data, then you have to disaggregate them to higher resolution, which is often not possible, or results in very crude estimates. Eventually, you have to make sure that your labelling of the industries is consistent with the Eurostat metadata vocabulary (“short codes”) or the iotables metadata vocabulary, i.e. you call the construction industry row CPA_F or construction.

kable (head(iotables::croatia_employment_2013, 5))

In case you use the Eurostat metadata vocabulary, your product / industry grouping variable may be called t_row2, or induse, for example. Make sure that you use the same variable name that is name of the key column in your use table. In this case it is .

Remember that we need to remove all totalling rows/columns and the extraterritorial organizations. Because we need a wide-format vector, and the employment data has only one dimension, you need to remove them only once: from the rows in case of the original table or from the columns if it is already spread out to wide form.

You need to add a matching key column in the beginning of the wide-form vector.

multiplier_name <- data.frame ( 
  iotables_row = "employment", stringsAsFactors = FALSE)

hr_emp <- croatia_employment_2013 %>%
  dplyr::select ( iotables_row, employment ) %>%
  tidyr::spread ( iotables_row, employment, !!2:ncol(.) )  %>%
  dplyr::select ( -total, -extraterriorial_organizations ) %>%  
  cbind(data.frame ( 
  iotables_row = "employment", stringsAsFactors = FALSE), .) 

kable (hr_emp[,1:5])

##Only the first 5 columns are shown to preserve space.

Now we can use the iotables functions to get the employment multiplier.

hr_emp_indicator <- input_indicator_create (
  input_matrix = hr_emp, 
  output_vector = select (output_vector_hr, -extraterriorial_organizations  )                        )

employment_multipliers_hr <- multiplier_create ( 
  input_vector    = hr_emp_indicator,
  Im              = I_hr, 
  multiplier_name = "employment_multiplier",
  digits          = 4 ) %>%
  tidyr::gather ( t_cols2, values, !! 2:ncol(.)) %>%
  mutate ( values = values * 1000 )%>%
  arrange (., desc(values))

kable(head(employment_multipliers_hr,23))

##Only the first 5 columns are shown to preserve space.