PKP 2026 - 02b Descriptive statistics
The structure and evolution of investment in Slovenia
Aljaz Colnar
2026-08-25
load("output/clean/00_setup.RData")
suppressPackageStartupMessages({ library(dplyr); library(ggplot2) })
dat <- readRDS("output/clean/pkp_clean.rds")
sections <- dat %>% filter(level == "section")
divisions <- dat %>% filter(level == "division")
si <- sections %>% filter(geo == "SI")# PERIODISATION: pre-crisis and post-crisis, split at 2008.
#
# The breakpoint follows the literature.
# Bontadini, Corrado, Haskel, Iommi & Jona-Lasinio (2023), the EUKLEMS & INTANProd
# methodology report (Deliverable D2.3.1), compare 1999-2007 with 2008-2019 when
# decomposing sources of growth for the EU and the US. The same break is standard
# in the EU-US intangibles literature.
#
# The endpoints are extended to 1995 and 2021 so that no observation is
# discarded but the breakpoint is theirs.
add_period <- function(d) {
d %>% mutate(period = case_when(
year <= 2007 ~ "1995-2007 pre-crisis",
TRUE ~ "2008-2021 post-crisis"
))
}
# Short labels for figure axes. These must be a FACTOR with explicit levels: as
# plain text ggplot would sort them alphabetically.
short_period <- c("1995-2007 pre-crisis" = "95-07",
"2008-2021 post-crisis" = "08-21")
period_order <- c("95-07", "08-21")
# Helper: convert a full period name to a correctly ordered short label
short_lab <- function(x) factor(unname(short_period[x]), levels = period_order)
# COUNTRY GROUPS - summary. The full reasoning is in 00_setup.Rmd, under
# "Why the country groups are what they are" (section 5), which also records the
# alternatives considered and the known weaknesses.
#
# Slovenia SI
# Nordic DK FI SE
# Continental AT DE FR LU NL
# Southern ES IT
# Central and Eastern BG CZ LT LV RO SK
# Non-EU UK US JP
#
# INCLUSION RULE. A country enters if BOTH tangible and intangible investment are
# present for at least 80 percent of the 162 possible observations (6 focus
# sections x 27 years). The bar is 80 rather than 100 because Slovenia itself is
# not complete - 162 tangible but 144 intangible, the gap being transport before
# 2006 - so requiring completeness would exclude comparators better covered than
# the subject of the study. Twenty-one countries qualify.
#
# Excluded, with the reason: BE (tangible only) · EL, HU, EE (intangible only) ·
# MT (tangible below threshold) · PT (both far below) · CY, HR, IE, PL (no
# investment data). All of these do have total-economy intangible investment in
# the `annual` table, so the restriction applies to the industry analysis only.
#
# THESE GROUPINGS ARE OUR OWN DECISION, NOT A LITERATURE STANDARD. The regional
# labels are conventional in European comparative work, but no source defines
# these memberships. What is sourced: the EU15 / new-member-state split that the
# four regions nest into, and the exclusions, which are driven by coverage. What
# is ours: which country sits in which region, six groups rather than two, and
# placing the United Kingdom outside the EU - which departs from Bontadini et al.
# (2023), whose EU aggregate includes it.
#
# The four EU regions nest exactly inside the standard split, so results can be
# reported at either level without recomputing:
# Nordic + Continental + Southern = EU15 members in the sample (10)
# Central and Eastern = post-2004 accession members (6)
geo_lookup <- data.frame(
geo = unlist(country_groups, use.names = FALSE),
peer_group = rep(names(country_groups), lengths(country_groups))
)
# group_names comes from 00_setup.Rmd, with its sources cited there.
# Cross-country and cross-period comparisons must use the SAME industries in
# every period, or a section entering the average partway through distorts the
# trend. Slovenia has no intangible investment data for transport (H) before
# 2006, and H's intangible intensity is far below the others, so including it
# only in later periods makes Slovenian intangible investment look flat when it
# in fact grew. H is therefore excluded from comparisons and reported separately.
consistent_sections <- c("C", "G", "I", "J", "M")1 Slovenia: how much is invested, and how much of it is intangible
Averages within each period, by section. All figures are percentages.
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(inv_intensity = round(mean(inv_intensity, na.rm = TRUE), 1),
tang_intensity = round(mean(tang_intensity, na.rm = TRUE), 1),
intang_intensity = round(mean(intang_intensity, na.rm = TRUE), 1),
intang_share = round(mean(intang_share, na.rm = TRUE), 1),
.groups = "drop") %>%
arrange(industry, period) %>%
as.data.frame()## industry period inv_intensity
## 1 Accommodation and food service 1995-2007 pre-crisis 27.9
## 2 Accommodation and food service 2008-2021 post-crisis 21.6
## 3 Information and communication 1995-2007 pre-crisis 39.7
## 4 Information and communication 2008-2021 post-crisis 25.6
## 5 Manufacturing 1995-2007 pre-crisis 25.9
## 6 Manufacturing 2008-2021 post-crisis 23.4
## 7 Professional, scientific and technical 1995-2007 pre-crisis 24.3
## 8 Professional, scientific and technical 2008-2021 post-crisis 17.6
## 9 Trade and repair of motor vehicles 1995-2007 pre-crisis 23.7
## 10 Trade and repair of motor vehicles 2008-2021 post-crisis 12.6
## 11 Transport and storage 1995-2007 pre-crisis 61.7
## 12 Transport and storage 2008-2021 post-crisis 30.6
## tang_intensity intang_intensity intang_share
## 1 26.2 5.0 16.6
## 2 20.2 5.3 23.8
## 3 32.7 11.9 27.6
## 4 14.5 17.4 55.2
## 5 19.4 12.7 39.7
## 6 15.7 14.6 48.3
## 7 13.3 24.0 64.4
## 8 7.5 22.3 75.0
## 9 20.6 10.1 32.8
## 10 10.4 10.9 51.9
## 11 56.1 4.1 5.7
## 12 28.8 4.5 15.1Columns. inv = gross fixed capital
formation ÷ gross value added, current prices — total investment on the
national accounts basis. tang = total
tangible investment ÷ adjusted gross value added. intang =
total intangible investment ÷ adjusted gross value added.
share = intangible ÷ (intangible + tangible). All
percentages.
!! inv uses a different numerator
and denominator from the other two, so
tang + intang does not equal
inv. inv excludes branding, organisational
capital and training, which national accounts treat as running
costs.
Tangible intensity falls in every section while intangible intensity barely moves, so the rising share is largely a denominator effect. This agrees with van Ark, de Vries & Erumban (Are Intangibles Running out of Steam?), who find for nine EU countries, the UK and the US that intangible capital deepening “has remained positive though not strong enough to offset the effects of the large decline in tangible capital deepening”. Slovenia looks like a version of the same pattern.
2 The sample, and how large each industry is
A reader needs to know how big each section is before interpreting its shares: a 46 percent rise in a small sector means something different from the same rise in a large one.
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(VAadj_mEUR = round(mean(VAadj, na.rm = TRUE)),
hours_m = round(mean(H_EMP, na.rm = TRUE) / 1000, 1),
persons_th = round(mean(EMP, na.rm = TRUE)),
.groups = "drop") %>%
group_by(period) %>%
mutate(pct_of_six = round(100 * VAadj_mEUR / sum(VAadj_mEUR), 1)) %>%
ungroup() %>%
arrange(period, desc(VAadj_mEUR)) %>%
as.data.frame()## industry period VAadj_mEUR
## 1 Manufacturing 1995-2007 pre-crisis 5003
## 2 Trade and repair of motor vehicles 1995-2007 pre-crisis 2375
## 3 Professional, scientific and technical 1995-2007 pre-crisis 1232
## 4 Transport and storage 1995-2007 pre-crisis 1034
## 5 Information and communication 1995-2007 pre-crisis 756
## 6 Accommodation and food service 1995-2007 pre-crisis 444
## 7 Manufacturing 2008-2021 post-crisis 8720
## 8 Trade and repair of motor vehicles 2008-2021 post-crisis 4730
## 9 Professional, scientific and technical 2008-2021 post-crisis 2837
## 10 Transport and storage 2008-2021 post-crisis 2246
## 11 Information and communication 2008-2021 post-crisis 1597
## 12 Accommodation and food service 2008-2021 post-crisis 819
## hours_m persons_th pct_of_six
## 1 402.3 247 46.1
## 2 184.5 112 21.9
## 3 74.8 44 11.4
## 4 78.1 47 9.5
## 5 28.4 17 7.0
## 6 46.7 30 4.1
## 7 329.6 205 41.6
## 8 193.7 121 22.6
## 9 121.6 73 13.5
## 10 85.8 52 10.7
## 11 46.4 28 7.6
## 12 55.8 37 3.9Columns. VAadj_mEUR = adjusted gross
value added, millions of euro. hours_m = million hours
worked. persons_th = thousand persons employed.
pct_of_six = the section’s share of the six focus sections
combined, so it is a share of the part of the economy studied here, not
of the whole economy.
The six focus sections differ by an order of magnitude in size, so unweighted averages across them - used throughout this document - give a small section the same weight as a large one. That is the right choice for describing industry behaviour, but it means these figures should not be read as economy-wide aggregates.
3 Summary statistics
The standard opening table: how many observations each indicator rests on, and the range it covers. Slovenia only, all six sections, full period.
key_indicators <- c("inv_intensity", "tang_intensity", "intang_intensity",
"intang_share", "ict_share",
"sh_compinfo", "sh_innovprop", "sh_econcomp", "sh_nonNA",
"lp_hour", "lp_person", "hours_per_person")
do.call(rbind, lapply(key_indicators, function(v) {
x <- si[[v]]
data.frame(indicator = v,
n = sum(!is.na(x)),
missing = sum(is.na(x)),
mean = round(mean(x, na.rm = TRUE), 2),
sd = round(sd(x, na.rm = TRUE), 2),
min = round(min(x, na.rm = TRUE), 2),
median = round(median(x, na.rm = TRUE), 2),
max = round(max(x, na.rm = TRUE), 2))
}))## indicator n missing mean sd min median max
## 1 inv_intensity 162 0 27.67 13.67 9.59 24.37 75.93
## 2 tang_intensity 162 0 21.90 13.95 5.00 18.41 72.58
## 3 intang_intensity 144 18 12.68 6.42 3.30 12.01 29.11
## 4 intang_share 144 18 41.14 19.60 5.46 39.63 81.81
## 5 ict_share 162 0 16.83 19.96 1.42 8.89 69.21
## 6 sh_compinfo 144 18 10.41 10.66 1.94 6.09 50.05
## 7 sh_innovprop 144 18 28.30 19.45 8.63 18.36 63.33
## 8 sh_econcomp 144 18 61.29 18.80 33.40 62.68 88.24
## 9 sh_nonNA 144 18 77.71 15.09 45.19 75.32 98.06
## 10 lp_hour 162 0 23.62 6.55 10.53 22.86 38.85
## 11 lp_person 162 0 38.51 10.96 16.44 37.99 64.30
## 12 hours_per_person 162 0 1627.86 81.31 1029.18 1634.10 1795.16Columns. n and missing out
of 162 (6 sections x 27 years). lp_hour is in euro
per hour, lp_person in thousand euro per
person, hours_per_person in hours per
year; every other row is a percentage.
4 Tangible investment: ICT against non-ICT (RQ1)
RQ1 asks for the split within tangible investment between information
and communications technology and everything else.
ict_share is computing equipment plus communications
equipment plus software, as a percentage of total gross fixed capital
formation.
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(ict_share = round(mean(ict_share, na.rm = TRUE), 1),
non_ict_share = round(100 - mean(ict_share, na.rm = TRUE), 1),
.groups = "drop") %>%
tidyr::pivot_wider(names_from = period, values_from = c(ict_share, non_ict_share)) %>%
as.data.frame()## industry ict_share_1995-2007 pre-crisis
## 1 Accommodation and food service 6.4
## 2 Information and communication 56.3
## 3 Manufacturing 6.8
## 4 Professional, scientific and technical 12.8
## 5 Trade and repair of motor vehicles 10.6
## 6 Transport and storage 3.0
## ict_share_2008-2021 post-crisis non_ict_share_1995-2007 pre-crisis
## 1 5.0 93.6
## 2 63.4 43.7
## 3 5.4 93.2
## 4 14.7 87.2
## 5 13.0 89.4
## 6 4.1 97.0
## non_ict_share_2008-2021 post-crisis
## 1 95.0
## 2 36.6
## 3 94.6
## 4 85.3
## 5 87.0
## 6 95.9Columns. Percentages of total gross fixed capital formation, national accounts basis. ICT = computing equipment + communications equipment + software and databases. The two columns sum to 100 by construction.
ICT capital is a small share of total investment in every section except information and communication, and the increase over time is modest compared with the shift toward intangibles reported above. That is consistent with the Europe’s Productivity Glass Half Full argument that ICT “is not a strong contributor to Total Factor Productivity Growth (”TFP”) growth in and of itself, because it mostly interacts with other aspects of digitisation” - the hardware is not where the action is.
5 Labour productivity
Real value added per hour and per person. Per hour is the primary measure, following EUKLEMS and OECD convention. (answer to prof. Redek question) Per person is also reported alongside
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(per_hour = round(mean(lp_hour, na.rm = TRUE), 1),
per_person = round(mean(lp_person, na.rm = TRUE), 1),
hours = round(mean(hours_per_person, na.rm = TRUE)),
.groups = "drop") %>%
arrange(industry, period) %>%
as.data.frame()## industry period per_hour
## 1 Accommodation and food service 1995-2007 pre-crisis 18.9
## 2 Accommodation and food service 2008-2021 post-crisis 16.5
## 3 Information and communication 1995-2007 pre-crisis 30.9
## 4 Information and communication 2008-2021 post-crisis 35.3
## 5 Manufacturing 1995-2007 pre-crisis 15.8
## 6 Manufacturing 2008-2021 post-crisis 27.4
## 7 Professional, scientific and technical 1995-2007 pre-crisis 26.1
## 8 Professional, scientific and technical 2008-2021 post-crisis 24.4
## 9 Trade and repair of motor vehicles 1995-2007 pre-crisis 18.1
## 10 Trade and repair of motor vehicles 2008-2021 post-crisis 24.4
## 11 Transport and storage 1995-2007 pre-crisis 18.2
## 12 Transport and storage 2008-2021 post-crisis 26.6
## per_person hours
## 1 29.5 1562
## 2 24.8 1507
## 3 51.0 1650
## 4 58.3 1650
## 5 25.7 1631
## 6 44.1 1612
## 7 44.1 1696
## 8 40.3 1662
## 9 29.9 1654
## 10 39.1 1607
## 11 30.3 1667
## 12 43.5 1643Columns. per_hour = adjusted real value
added per hour worked, euro at 2015 prices.
per_person = the same per person employed, thousand
euro. hours = average hours worked per person per
year. The three carry different units because the source variables do:
value added is in millions of euro, hours and persons in thousands.
Within-country comparison over time only - these are in national
currency and cannot be compared across countries.
# Growth in real labour productivity per hour, pre-crisis to post-crisis
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(lp = mean(lp_hour, na.rm = TRUE), .groups = "drop") %>%
group_by(industry) %>%
summarise(pre_crisis = round(first(lp), 1),
post_crisis = round(last(lp), 1),
change_pct = round(100 * (last(lp) / first(lp) - 1), 1),
.groups = "drop") %>%
arrange(desc(change_pct)) %>%
as.data.frame()## industry pre_crisis post_crisis change_pct
## 1 Manufacturing 15.8 27.4 73.7
## 2 Transport and storage 18.2 26.6 46.3
## 3 Trade and repair of motor vehicles 18.1 24.4 35.2
## 4 Information and communication 30.9 35.3 14.1
## 5 Professional, scientific and technical 26.1 24.4 -6.5
## 6 Accommodation and food service 18.9 16.5 -12.5This is the variable the whole intangibles literature exists to explain, so the ordering matters: the sections that raised intangible intensity most should, on the Corrado-Hulten-Sichel account, show the stronger productivity growth. Van Ark, de Vries & Erumban find the opposite at the frontier - “a relatively strong slowdown in labour productivity growth for the most intangible-intensive industries” in the UK and US. Whether Slovenia follows the theory or the recent frontier evidence is a genuine question, and the growth-accounting section (RQ4) is where it should be answered properly rather than inferred from these two columns.
6 The evolution of the intangible share
The headline indicator, by section, at five-year intervals.
si %>%
filter(year %in% c(1995, 2000, 2005, 2010, 2015, 2021)) %>%
select(industry, year, intang_share) %>%
mutate(intang_share = round(intang_share, 1)) %>%
tidyr::pivot_wider(names_from = year, values_from = intang_share) %>%
as.data.frame()## industry 1995 2000 2005 2010 2015 2021
## 1 Manufacturing 43.7 38.5 39.7 54.2 50.1 45.6
## 2 Trade and repair of motor vehicles NA 28.0 35.5 47.4 54.1 56.2
## 3 Transport and storage NA NA NA 14.6 16.3 11.1
## 4 Accommodation and food service 21.9 17.2 14.0 15.6 32.0 NA
## 5 Information and communication 27.3 19.9 35.9 56.2 55.4 57.4
## 6 Professional, scientific and technical 69.2 58.4 57.6 74.8 77.7 NAColumns. Years across the top. Values are
intang_share = intangible investment ÷ (intangible +
tangible investment), percent. A dash means no data for that year.
6.0.1 Figure
si %>%
filter(!is.na(intang_share)) %>%
ggplot(aes(year, intang_share)) +
geom_hline(yintercept = 50, linetype = "dotted", colour = "grey40") +
geom_line(linewidth = 0.6) +
facet_wrap(~ industry, ncol = 3) +
labs(title = "Intangible share of total investment, Slovenia, 1995-2021",
subtitle = "percent; dotted line marks the point where intangible exceeds tangible",
x = NULL, y = "percent") +
theme_pkp()
Figure 1. Intangible investment as a share of total investment, by industry, Slovenia, 1995-2021, percent. Source: EUKLEMS & INTANProd, own calculations.
# In which year did intangible investment first exceed tangible?
si %>%
filter(!is.na(intang_share)) %>%
group_by(industry) %>%
summarise(first_year_above_50 = ifelse(any(intang_share > 50),
min(year[intang_share > 50]),
NA_integer_),
share_1995 = round(first(intang_share[order(year)]), 1),
share_2021 = round(last(intang_share[order(year)]), 1),
.groups = "drop") %>%
as.data.frame()## industry first_year_above_50 share_1995
## 1 Accommodation and food service NA 21.9
## 2 Information and communication 2009 27.3
## 3 Manufacturing 2010 43.7
## 4 Professional, scientific and technical 1995 69.2
## 5 Trade and repair of motor vehicles 2012 33.8
## 6 Transport and storage NA 5.9
## share_2021
## 1 20.6
## 2 57.4
## 3 45.6
## 4 78.4
## 5 56.2
## 6 11.1Columns. first_year_above_50 = the
first year intangible investment exceeded tangible.
share_1995 and share_2021 are the first and
last available years, which for some industries is not 1995 or 2021 —
transport begins in 2006.
Three sections cross 50 percent within four years of the crisis, and manufacturing then falls back below. That timing is consistent with the crisis-centred break used throughout this literature, and with van Ark et al.’s attribution of the shift to collapsing tangible investment rather than accelerating intangible investment.
7 What kind of intangible investment?
The three Corrado-Hulten-Sichel categories, as shares of total intangible investment. They sum to 100.
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(computerised_info = round(mean(sh_compinfo, na.rm = TRUE), 1),
innovative_property = round(mean(sh_innovprop, na.rm = TRUE), 1),
economic_competencies= round(mean(sh_econcomp, na.rm = TRUE), 1),
.groups = "drop") %>%
arrange(industry, period) %>%
as.data.frame()## industry period
## 1 Accommodation and food service 1995-2007 pre-crisis
## 2 Accommodation and food service 2008-2021 post-crisis
## 3 Information and communication 1995-2007 pre-crisis
## 4 Information and communication 2008-2021 post-crisis
## 5 Manufacturing 1995-2007 pre-crisis
## 6 Manufacturing 2008-2021 post-crisis
## 7 Professional, scientific and technical 1995-2007 pre-crisis
## 8 Professional, scientific and technical 2008-2021 post-crisis
## 9 Trade and repair of motor vehicles 1995-2007 pre-crisis
## 10 Trade and repair of motor vehicles 2008-2021 post-crisis
## 11 Transport and storage 1995-2007 pre-crisis
## 12 Transport and storage 2008-2021 post-crisis
## computerised_info innovative_property economic_competencies
## 1 4.0 9.7 86.3
## 2 4.5 14.4 81.1
## 3 21.3 12.2 66.4
## 4 37.5 16.0 46.5
## 5 4.7 47.8 47.5
## 6 5.1 51.8 43.0
## 7 4.8 58.5 36.7
## 8 6.4 53.9 39.7
## 9 5.8 9.2 85.0
## 10 6.7 14.4 78.9
## 11 8.7 20.2 71.1
## 12 12.1 20.5 67.4Columns. The three Corrado-Hulten-Sichel categories
as percentages of total intangible investment; each row
sums to 100. computerised_info = software and databases.
innovative_property = R&D, industrial design, new
financial products, artistic originals, mineral exploration.
economic_competencies = brand, organisational capital,
training.
Economic competencies dominate in trade, transport and accommodation (66-87 percent) but not in manufacturing or professional services (36-49). Corrado, Haskel, Jona-Lasinio & Iommi (2022) report that the correlation between growth in R&D capital and in non-R&D intangible capital is only 0.32 - the two kinds of intangible move largely independently.
7.0.1 Figure
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(`Computerised information` = mean(sh_compinfo, na.rm = TRUE),
`Innovative property` = mean(sh_innovprop, na.rm = TRUE),
`Economic competencies` = mean(sh_econcomp, na.rm = TRUE),
.groups = "drop") %>%
tidyr::pivot_longer(-c(industry, period), names_to = "category", values_to = "pct") %>%
filter(!is.na(pct)) %>%
mutate(period = short_lab(period),
category = factor(category, levels = c("Computerised information",
"Innovative property",
"Economic competencies"))) %>%
ggplot(aes(period, pct, fill = category)) +
geom_col(width = 0.7, colour = "black", linewidth = 0.2) +
facet_wrap(~ industry, ncol = 3) +
scale_fill_grey(start = 0.25, end = 0.9) +
labs(title = "Composition of intangible investment, Slovenia",
subtitle = "the three Corrado-Hulten-Sichel categories, percent of total intangible investment",
x = NULL, y = "percent") +
theme_pkp()
Figure 2. Composition of intangible investment by Corrado-Hulten-Sichel category, Slovenia, by period, percent. Source: EUKLEMS & INTANProd, own calculations. # Question for professors: Not so much insight - should we make more time intervals ex. 93-04 (before EU), 04-14 (before crisis and crisis), after 14 (after crisis) or should we follow literature?
7.1 Individual assets
A finer breakdown of the same total. These sit inside the three categories above, so do not add them to those.
si %>%
filter(year >= 2008) %>%
group_by(industry) %>%
summarise(software = round(mean(sh_compinfo, na.rm = TRUE), 1),
rd = round(mean(sh_rd, na.rm = TRUE), 1),
design = round(mean(sh_design, na.rm = TRUE), 1),
branding = round(mean(sh_brand, na.rm = TRUE), 1),
orgcap = round(mean(sh_orgcap, na.rm = TRUE), 1),
training = round(mean(sh_train, na.rm = TRUE), 1),
.groups = "drop") %>%
as.data.frame()## industry software rd design branding orgcap
## 1 Accommodation and food service 4.5 0.2 14.2 33.2 37.2
## 2 Information and communication 37.5 9.1 7.0 22.4 21.2
## 3 Manufacturing 5.1 32.1 18.6 15.8 18.9
## 4 Professional, scientific and technical 6.4 26.8 27.1 15.1 19.7
## 5 Trade and repair of motor vehicles 6.7 1.4 13.0 40.5 30.4
## 6 Transport and storage 12.1 0.9 19.6 13.6 41.1
## training
## 1 10.6
## 2 2.9
## 3 8.2
## 4 4.9
## 5 8.0
## 6 12.7Columns. Individual assets as percentages of total intangible investment.
Attention! These sit inside the three categories above, not alongside them — R&D and design are part of innovative property; branding, organisational capital and training are part of economic competencies. Do not add both levels together.
R&D is close to absent from Slovenian services - 1.3 percent of intangible investment in trade, 0.5 in transport, 0.2 in accommodation - and only 7.0 percent even in ICT, whose intangible investment is overwhelmingly software. Read against Corrado et al. (2022), this matters for interpretation: a chapter that treated R&D as the proxy for knowledge investment would conclude Slovenian services invest in nothing, when on the broader measure they invest substantially.
8 The measurement gap
The share of intangible investment that national accounts do not capitalise - the gap Corrado, Hulten and Sichel wrote about, measured for Slovenian industries.
si %>%
add_period() %>%
group_by(industry, period) %>%
summarise(not_in_national_accounts = round(mean(sh_nonNA, na.rm = TRUE), 1),
.groups = "drop") %>%
tidyr::pivot_wider(names_from = period, values_from = not_in_national_accounts) %>%
as.data.frame()## industry 1995-2007 pre-crisis
## 1 Accommodation and food service 96.0
## 2 Information and communication 75.9
## 3 Manufacturing 65.6
## 4 Professional, scientific and technical 72.0
## 5 Trade and repair of motor vehicles 93.2
## 6 Transport and storage 89.5
## 2008-2021 post-crisis
## 1 95.3
## 2 53.5
## 3 61.7
## 4 66.8
## 5 91.9
## 6 87.0Columns. Share of intangible investment that
national accounts do not capitalise:
I_NonNatAcc ÷ I_Intang, percent. Software, R&D and
other intellectual property are capitalised; branding, organisational
capital, training and design are not. A value of 90 means nine-tenths of
what the industry invests in knowledge is invisible in official
investment statistics.
The gap ranges from 50.5 percent in ICT to 95.2 in accommodation. Corrado et al. (2022) warn that where intangible investment is uncounted, “the higher the (uncounted) intangible investment, the greater the misperception”. Because the gap here is systematically larger in traditional services, official statistics do not understate Slovenian industries uniformly - they understate some far more than others, which biases any ranking built on national accounts.
9 Is Slovenia’s measurement gap larger than its comparators’?
Official statistics capitalise software, research and development, and other intellectual property, but not branding, organisational capital, training or design. If Slovenian services rely more heavily on the uncounted kinds of intangible than German or Dutch services do, then official figures understate Slovenia by more than they understate its comparators - and part of the investment gap between Slovenia and the EU would be a measurement artefact rather than a real difference.
The five-section set is used again, so the comparison is apples with apples.
gap <- sections %>%
inner_join(geo_lookup, by = "geo") %>%
filter(nace %in% consistent_sections, year >= 2008) %>%
mutate(peer_group = factor(unname(group_names[peer_group]),
levels = unname(group_names)))
gap %>%
group_by(peer_group) %>%
summarise(not_capitalised_pct = round(mean(sh_nonNA, na.rm = TRUE), 1),
econ_competencies = round(mean(sh_econcomp, na.rm = TRUE), 1),
.groups = "drop") %>%
as.data.frame()## peer_group not_capitalised_pct econ_competencies
## 1 Slovenia 73.6 57.8
## 2 Nordic 63.3 53.5
## 3 Continental 64.4 57.2
## 4 Southern 62.3 50.1
## 5 Central and Eastern 78.8 67.9
## 6 Non-EU 61.9 52.9Columns. not_capitalised_pct = share of
intangible investment outside national accounts.
econ_competencies = share that is brand, organisational
capital or training. The first is always larger, because it also
includes design and new financial products.
9.1 Slovenia against each comparator, industry by industry
This table shows the gap (the percentage of that industry’s intangible investment that national accounts do not count as investment) for every industry and group.
gap %>%
group_by(industry, peer_group) %>%
summarise(g = round(mean(sh_nonNA, na.rm = TRUE), 1), .groups = "drop") %>%
tidyr::pivot_wider(names_from = peer_group, values_from = g) %>%
as.data.frame()## industry Slovenia Nordic Continental Southern
## 1 Accommodation and food service 95.3 90.1 88.4 86.8
## 2 Information and communication 53.5 43.2 45.3 35.9
## 3 Manufacturing 61.7 44.8 45.4 57.4
## 4 Professional, scientific and technical 66.8 61.0 69.4 50.9
## 5 Trade and repair of motor vehicles 91.9 77.4 73.3 81.1
## Central and Eastern Non-EU
## 1 91.9 83.5
## 2 52.5 48.2
## 3 78.8 40.5
## 4 82.2 61.0
## 5 90.1 75.8Columns. The same measurement gap, split by industry and peer group, so a difference concentrated in one sector is visible rather than averaged away.
10 Knowledge-intensive against traditional services (RQ3)
J and M against G, H and I. Manufacturing shown for reference.
si %>%
add_period() %>%
group_by(group, period) %>%
summarise(intang_share = round(mean(intang_share, na.rm = TRUE), 1),
intang_intensity = round(mean(intang_intensity, na.rm = TRUE), 1),
software_share = round(mean(sh_compinfo, na.rm = TRUE), 1),
econcomp_share = round(mean(sh_econcomp, na.rm = TRUE), 1),
.groups = "drop") %>%
arrange(group, period) %>%
as.data.frame()## group period intang_share intang_intensity
## 1 Knowledge-intensive 1995-2007 pre-crisis 46.0 18.0
## 2 Knowledge-intensive 2008-2021 post-crisis 64.7 19.8
## 3 Manufacturing 1995-2007 pre-crisis 39.7 12.7
## 4 Manufacturing 2008-2021 post-crisis 48.3 14.6
## 5 Traditional 1995-2007 pre-crisis 22.6 7.1
## 6 Traditional 2008-2021 post-crisis 31.6 7.1
## software_share econcomp_share
## 1 13.1 51.5
## 2 22.5 43.2
## 3 4.7 47.5
## 4 5.1 43.0
## 5 5.1 84.6
## 6 7.5 76.3Columns. intang_share = intangible ÷
all investment. intang_intensity = intangible investment ÷
adjusted value added. software_share and
econcomp_share are percentages of total intangible
investment. Groups: knowledge-intensive = J and M; traditional = G, H
and I.
Knowledge-intensive services raise intangible intensity (18.0 to 19.8) while traditional services do not move at all (7.1 to 7.1), so the two groups diverge rather than converge. Van Ark et al. find “a relatively strong slowdown in labour productivity growth for the most intangible-intensive industries” in the UK and US - the opposite direction from Slovenia, where the intangible-intensive group is the only one still increasing. Whether that reflects catch-up from a lower base or something specific to Slovenia is a question for the growth-accounting section.
10.0.1 Figure
The two panels make the distinction that the table encodes but does not emphasise. On the share, all three groups rise. On the intensity - investment relative to output - traditional services do not move at all.
grp_fig <- si %>%
add_period() %>%
group_by(group, period) %>%
summarise(`Intangible intensity (% of value added)` =
mean(intang_intensity, na.rm = TRUE),
`Intangible share (% of all investment)` =
mean(intang_share, na.rm = TRUE),
.groups = "drop") %>%
tidyr::pivot_longer(-c(group, period), names_to = "measure", values_to = "pct") %>%
mutate(period = short_lab(period),
group = factor(group, levels = c("Knowledge-intensive",
"Traditional", "Manufacturing")))
fig_groups <- ggplot(grp_fig, aes(group, pct, fill = period)) +
geom_col(position = position_dodge(width = 0.75), width = 0.65,
colour = "black", linewidth = 0.2) +
facet_wrap(~ measure, scales = "free_y") +
scale_fill_grey(start = 0.75, end = 0.35) +
labs(title = "Knowledge-intensive against traditional services, Slovenia",
subtitle = "the share rises in all three groups; the intensity does not",
x = NULL, y = "percent") +
theme_pkp()
fig_groups
Figure 3. Intangible investment intensity and intangible share of total investment, by industry group, Slovenia, pre-crisis and post-crisis. Groups: knowledge-intensive = J and M; traditional = G, H and I. Source: EUKLEMS & INTANProd, own calculations.
11 Slovenia against the peer groups (RQ2 groundwork)
11.1 Holding the industry set constant
Comparisons over time must use the same industries in every period, or a section entering the average partway through distorts the trend.
This is the challange here. Slovenia has no intangible investment data for transport (H) before 2006, so H is absent from Slovenia’s early averages and present in the later ones - and its intangible intensity is only about 4.5 percent, far below the others. Including it in the last period alone makes Slovenia’s intangible investment look flat when it in fact grew.
Everything in this section therefore uses a fixed set of five sections, excluding transport. Transport is reported separately.
# consistent_sections, geo_lookup and group_names are defined at the top.
peers <- sections %>%
inner_join(geo_lookup, by = "geo") %>%
filter(nace %in% consistent_sections) %>%
add_period() %>%
mutate(peer_group = factor(unname(group_names[peer_group]),
levels = unname(group_names)),
# coarser aggregation, for reporting at either level
bloc = case_when(geo %in% eu15_comparators ~ "EU15",
geo %in% nms_comparators ~ "New member states",
geo %in% non_eu ~ "Non-EU",
geo == "SI" ~ "Slovenia"),
bloc = factor(bloc, levels = c("Slovenia", "New member states",
"EU15", "Non-EU")))
# Confirm the set really is balanced
peers %>%
count(peer_group, period, name = "observations") %>%
tidyr::pivot_wider(names_from = period, values_from = observations) %>%
as.data.frame()## peer_group 1995-2007 pre-crisis 2008-2021 post-crisis
## 1 Slovenia 65 70
## 2 Nordic 195 210
## 3 Continental 325 350
## 4 Southern 130 140
## 5 Central and Eastern 390 420
## 6 Non-EU 195 210Columns. Observation counts per group and period (countries × 5 sections × years in period). They must be equal across periods, or a section entering the average partway through would distort the trend.
11.2 An empirical “intangible frontier”
The background paper names Belgium, Ireland, Denmark, Luxembourg and the Netherlands as leading the EU in intangible investment intensity. Belgium and Ireland have no intangible investment data in EUKLEMS and Luxembourg is not in the focus set, so that group cannot be reproduced here.
Rather than substituting countries by hand - which would be our own construction presented as the literature’s - the frontier is defined by a rule: the countries in the top quartile of intangible investment intensity in the pre-crisis period. It is computed from the data, so it can be reproduced and defended, and it is reported alongside the sourced groups rather than replacing them.
frontier_countries <- sections %>%
filter(nace %in% consistent_sections, year <= 2007,
geo %in% peer_countries) %>%
group_by(geo) %>%
summarise(pre_intensity = mean(intang_intensity, na.rm = TRUE), .groups = "drop") %>%
filter(pre_intensity >= quantile(pre_intensity, 0.75, na.rm = TRUE)) %>%
arrange(desc(pre_intensity))
frontier_countries## # A tibble: 5 × 2
## geo pre_intensity
## <chr> <dbl>
## 1 SE 17.5
## 2 US 16.4
## 3 FR 16.3
## 4 FI 16.3
## 5 UK 15.8Columns. pre_intensity = mean
intangible investment as a percentage of adjusted gross value added,
1995-2007, across the five consistent sections. Countries at or above
the 75th percentile are shown.
11.3 At the coarser level: Slovenia, new member states, EU15, non-EU
The four regional groups nest inside the standard EU15 / new-member-state split, so the same comparison can be read at either level. This is the coarser one.
peers %>%
group_by(bloc, period) %>%
summarise(tangible = round(mean(tang_intensity, na.rm = TRUE), 1),
intangible = round(mean(intang_intensity, na.rm = TRUE), 1),
share = round(mean(intang_share, na.rm = TRUE), 1),
n_countries = n_distinct(geo),
.groups = "drop") %>%
as.data.frame()## bloc period tangible intangible share n_countries
## 1 Slovenia 1995-2007 pre-crisis 22.4 12.8 36.3 1
## 2 Slovenia 2008-2021 post-crisis 13.7 14.1 50.9 1
## 3 New member states 1995-2007 pre-crisis 20.9 9.8 34.4 6
## 4 New member states 2008-2021 post-crisis 13.2 11.8 46.3 6
## 5 EU15 1995-2007 pre-crisis 11.1 12.4 50.3 10
## 6 EU15 2008-2021 post-crisis 8.9 14.9 58.8 10
## 7 Non-EU 1995-2007 pre-crisis 11.1 14.3 55.3 3
## 8 Non-EU 2008-2021 post-crisis 8.4 15.6 63.2 3Columns. As above. n_countries is shown
because the blocs differ in size - ten EU15 members against six new
member states and three non-EU comparators - and these are unweighted
means of country-level ratios.
11.4 Tangible against intangible, over time
The table that matters. Investment intensity is investment as a
percentage of adjusted gross value added
(VAadj), so tangible and intangible are directly
comparable.
peers %>%
group_by(peer_group, period) %>%
summarise(tangible = round(mean(tang_intensity, na.rm = TRUE), 1),
intangible = round(mean(intang_intensity, na.rm = TRUE), 1),
share = round(mean(intang_share, na.rm = TRUE), 1),
.groups = "drop") %>%
arrange(peer_group, period) %>%
as.data.frame()## peer_group period tangible intangible share
## 1 Slovenia 1995-2007 pre-crisis 22.4 12.8 36.3
## 2 Slovenia 2008-2021 post-crisis 13.7 14.1 50.9
## 3 Nordic 1995-2007 pre-crisis 10.6 16.2 57.8
## 4 Nordic 2008-2021 post-crisis 7.8 18.6 68.0
## 5 Continental 1995-2007 pre-crisis 11.0 11.0 48.7
## 6 Continental 2008-2021 post-crisis 9.3 13.5 55.9
## 7 Southern 1995-2007 pre-crisis 12.0 10.1 43.0
## 8 Southern 2008-2021 post-crisis 9.7 12.9 52.1
## 9 Central and Eastern 1995-2007 pre-crisis 20.9 9.8 34.4
## 10 Central and Eastern 2008-2021 post-crisis 13.2 11.8 46.3
## 11 Non-EU 1995-2007 pre-crisis 11.1 14.3 55.3
## 12 Non-EU 2008-2021 post-crisis 8.4 15.6 63.2Columns. tangible and
intangible are investment as a percentage of adjusted gross
value added — same denominator, directly comparable. share
= intangible ÷ all investment, computed as the unweighted mean of
individual country-industry-year shares. ⚠️ Because it is a mean of
ratios rather than a ratio of means, it will not exactly equal
intangible ÷ (tangible + intangible) from the two
columns.
11.4.1 Figure
peers %>%
group_by(peer_group, period) %>%
summarise(Tangible = mean(tang_intensity, na.rm = TRUE),
Intangible = mean(intang_intensity, na.rm = TRUE),
.groups = "drop") %>%
tidyr::pivot_longer(c(Tangible, Intangible), names_to = "type", values_to = "pct") %>%
mutate(period = short_lab(period)) %>%
ggplot(aes(period, pct, group = type, linetype = type, shape = type)) +
geom_line(linewidth = 0.6) +
geom_point(size = 1.8, fill = "white") +
facet_wrap(~ peer_group, ncol = 3) +
scale_shape_manual(values = c(21, 16)) +
labs(title = "Tangible and intangible investment intensity",
subtitle = "percent of adjusted gross value added; five sections held constant",
x = NULL, y = "percent") +
theme_pkp()
Figure 3. Tangible and intangible investment intensity by peer group, percent of adjusted gross value added, five sections (C, G, I, J, M). Source: EUKLEMS & INTANProd, own calculations.
# Change from the first period to the last, in percent
peers %>%
group_by(peer_group, period) %>%
summarise(tang = mean(tang_intensity, na.rm = TRUE),
int = mean(intang_intensity, na.rm = TRUE), .groups = "drop") %>%
group_by(peer_group) %>%
summarise(tangible_change_pct = round(100 * (last(tang) / first(tang) - 1), 1),
intangible_change_pct = round(100 * (last(int) / first(int) - 1), 1),
.groups = "drop") %>%
as.data.frame()## peer_group tangible_change_pct intangible_change_pct
## 1 Slovenia -39.1 9.9
## 2 Nordic -26.5 14.7
## 3 Continental -15.8 22.1
## 4 Southern -18.7 28.1
## 5 Central and Eastern -36.7 21.3
## 6 Non-EU -24.5 9.2Columns. Percentage change in each intensity from the pre-crisis period (1995-2007) to the post-crisis period (2008-2021). Negative means the intensity fell.
Tangible intensity falls in every group - between 12 and 39 percent - while intangible intensity rises everywhere. That is the pattern van Ark et al. document for the EU, the UK and the US, and it holds here across all six groups, which supports treating it as a general shift rather than a Slovenian story. Slovenia is at the extreme of both: the largest tangible fall and among the smallest intangible rises.
11.5 Is the pattern uniform, or driven by one or two industries?
Pooling five sections into one average can hide a lot. This repeats the change calculation industry by industry.
peers %>%
group_by(peer_group, industry, period) %>%
summarise(int = mean(intang_intensity, na.rm = TRUE), .groups = "drop") %>%
group_by(peer_group, industry) %>%
summarise(intangible_change_pct = round(100 * (last(int) / first(int) - 1), 1),
.groups = "drop") %>%
tidyr::pivot_wider(names_from = peer_group, values_from = intangible_change_pct) %>%
as.data.frame()## industry Slovenia Nordic Continental Southern
## 1 Accommodation and food service 5.5 36.5 10.3 18.3
## 2 Information and communication 46.0 17.6 28.6 38.0
## 3 Manufacturing 14.9 22.0 22.9 24.6
## 4 Professional, scientific and technical -7.2 1.1 20.5 26.8
## 5 Trade and repair of motor vehicles 8.0 16.9 19.3 13.5
## Central and Eastern Non-EU
## 1 13.4 8.0
## 2 39.3 1.9
## 3 10.9 16.0
## 4 6.4 6.8
## 5 32.1 19.2Columns. Percentage change in intangible investment intensity, first period to last, by industry and peer group. Shows whether a group’s overall figure is uniform or driven by one or two sectors.
Slovenian ICT grew intangible intensity by 46.0 percent, more than any group, while the non-EU comparators (UK, US, JP) grew just 1.9 percent. Van Ark et al. report exactly that stagnation for the most intangible-intensive industries in the UK and US. The two findings fit together: the frontier has slowed while a catching-up economy has not - which is what convergence looks like, and is worth testing directly in the convergence analysis (RQ5).
11.5.1 Figure
peers %>%
group_by(peer_group, industry, period) %>%
summarise(i = mean(intang_intensity, na.rm = TRUE), .groups = "drop") %>%
group_by(peer_group, industry) %>%
summarise(change = 100 * (last(i) / first(i) - 1), .groups = "drop") %>%
ggplot(aes(industry, change)) +
geom_hline(yintercept = 0, colour = "black", linewidth = 0.3) +
geom_col(width = 0.7, fill = "grey55", colour = "black", linewidth = 0.2) +
coord_flip() +
facet_wrap(~ peer_group, ncol = 3) +
labs(title = "Change in intangible investment intensity, 1995-2007 to 2008-2021",
subtitle = "percent change; bars below zero indicate a fall",
x = NULL, y = "percent change") +
theme_pkp()
Figure 4. Percentage change in intangible investment intensity between 1995-2007 and 2008-2021, by industry and peer group. Source: EUKLEMS & INTANProd, own calculations.
11.6 Professional services: why the intensity fell
Professional and scientific services is the only industry, in any peer group, where intangible investment intensity fell between the pre-crisis and post-crisis periods, while it rose in every other peer group. Because it was the largest contributor to Slovenia’s average, that fall drags the pooled national figure down.
Fherefore it needs checking :)
11.6.1 Checking starting level
m <- sections %>%
inner_join(geo_lookup, by = "geo") %>%
filter(nace == "M", year <= 2007)
m %>%
group_by(geo) %>%
summarise(intangible_intensity = round(mean(intang_intensity, na.rm = TRUE), 1),
.groups = "drop") %>%
arrange(desc(intangible_intensity)) %>%
as.data.frame()## geo intangible_intensity
## 1 SE 29.3
## 2 FR 28.8
## 3 CZ 24.8
## 4 FI 24.7
## 5 SI 24.0
## 6 DK 21.3
## 7 UK 19.9
## 8 NL 19.3
## 9 BG 19.2
## 10 LV 19.1
## 11 ES 16.5
## 12 RO 16.2
## 13 SK 15.9
## 14 AT 15.7
## 15 IT 13.7
## 16 LT 13.4
## 17 US 12.9
## 18 DE 11.8
## 19 LU 10.5
## 20 JP NaNColumns. Intangible investment as a percentage of adjusted gross value added, professional services, 1995-2007 (the pre-crisis period), by country.
Result. Slovenia is third of fifteen, between France and Czechia and in the same band as Sweden and Finland. The starting level is high but normal.
11.6.2 Is the decline smooth, or driven by odd years?
sections %>%
filter(geo == "SI", nace == "M", year <= 2005) %>%
transmute(year,
I_Intang = round(I_Intang, 1),
I_Tang = round(I_Tang, 1),
VAadj = round(VAadj, 1),
intensity = round(intang_intensity, 1)) %>%
as.data.frame()## year I_Intang I_Tang VAadj intensity
## 1 1995 152.7 67.9 524.6 29.1
## 2 1996 173.7 72.7 601.6 28.9
## 3 1997 174.8 97.3 647.1 27.0
## 4 1998 200.5 103.1 757.0 26.5
## 5 1999 218.2 112.0 904.0 24.1
## 6 2000 237.5 169.4 950.0 25.0
## 7 2001 273.7 185.0 1086.8 25.2
## 8 2002 299.7 170.7 1403.7 21.4
## 9 2003 328.8 146.6 1561.1 21.1
## 10 2004 363.4 208.9 1673.4 21.7
## 11 2005 364.5 268.5 1760.7 20.7Columns. I_Intang = intangible
investment, I_Tang = tangible investment,
VAadj = adjusted gross value added, all in millions of
euro. intensity = I_Intang / VAadj,
percent.
11.6.3 Figure
sections %>%
filter(geo == "SI", nace == "M") %>%
arrange(year) %>%
mutate(`Intangible investment` = 100 * I_Intang / first(I_Intang),
`Adjusted gross value added`= 100 * VAadj / first(VAadj)) %>%
tidyr::pivot_longer(c(`Intangible investment`, `Adjusted gross value added`),
names_to = "series", values_to = "index") %>%
ggplot(aes(year, index, linetype = series)) +
geom_line(linewidth = 0.7) +
labs(title = "Professional services: investment grew, output grew faster",
subtitle = "index, 1995 = 100. The falling intensity is a denominator effect",
x = NULL, y = "index, 1995 = 100") +
theme_pkp()
Figure 5. Intangible investment and adjusted gross value added, professional, scientific and technical activities, Slovenia, index 1995 = 100. Source: EUKLEMS & INTANProd, own calculations.
11.6.4 What the decline actually is
The series is smooth - no spikes, no missing years, no jumps. And the reason for the fall is in the two level columns:
1995 to 2005
intangible investment 152.7 -> 364.5 +139 percent
adjusted value added 524.6 -> 1760.7 +236 percentIntangible investment more than doubled. Output more than tripled. The intensity fell because the denominator grew faster than the numerator, not because anyone invested less.
This matters for how the result is written up. “Professional services went backwards” would be wrong. The sector expanded very fast in the decade after independence, and its intangible investment - though growing strongly in absolute terms - did not keep pace. That is a transition-economy pattern: a sector building itself out at speed, with knowledge investment following rather than leading.
Note also that the decline is concentrated in 1995-2005 (29.1 down to 20.7). The post-crisis average is 22.3, slightly above the 2005 level, so the intensity fell during the rapid expansion and then stabilised.
11.7 Transport, reported separately
Excluded above because Slovenia has no intangible investment data before 2006.
sections %>%
inner_join(geo_lookup, by = "geo") %>%
filter(nace == "H") %>%
add_period() %>%
group_by(peer_group, period) %>%
summarise(n_obs = sum(!is.na(intang_intensity)),
tangible = round(mean(tang_intensity, na.rm = TRUE), 1),
intangible = round(mean(intang_intensity, na.rm = TRUE), 1),
.groups = "drop") %>%
as.data.frame()## peer_group period n_obs tangible intangible
## 1 cee 1995-2007 pre-crisis 72 28.8 4.5
## 2 cee 2008-2021 post-crisis 80 28.7 5.4
## 3 continental 1995-2007 pre-crisis 65 28.5 6.1
## 4 continental 2008-2021 post-crisis 70 26.7 6.6
## 5 non_eu 1995-2007 pre-crisis 39 24.3 6.4
## 6 non_eu 2008-2021 post-crisis 42 23.3 7.0
## 7 nordic 1995-2007 pre-crisis 39 30.1 5.1
## 8 nordic 2008-2021 post-crisis 42 30.2 7.0
## 9 slovenia 1995-2007 pre-crisis 2 56.1 4.1
## 10 slovenia 2008-2021 post-crisis 11 28.8 4.5
## 11 southern 1995-2007 pre-crisis 26 36.8 4.2
## 12 southern 2008-2021 post-crisis 28 29.9 5.1Columns. Section H alone, excluded from the
comparisons above because Slovenia has no intangible investment data
before 2006. n_obs shows how many observations the averages
rest on .
11.8 Where Slovenia stands now
Unweighted means of country-level ratios, since all monetary values are in national currency and cannot be summed across countries.
sections %>%
inner_join(geo_lookup, by = "geo") %>%
filter(year >= 2008) %>%
group_by(peer_group, industry) %>%
summarise(intang_share = round(mean(intang_share, na.rm = TRUE), 1),
.groups = "drop") %>%
tidyr::pivot_wider(names_from = peer_group, values_from = intang_share) %>%
as.data.frame()## industry cee continental non_eu nordic
## 1 Accommodation and food service 21.0 32.0 47.4 47.5
## 2 Information and communication 59.9 58.1 68.3 72.7
## 3 Manufacturing 34.5 56.3 65.9 70.2
## 4 Professional, scientific and technical 69.1 81.7 83.3 86.0
## 5 Trade and repair of motor vehicles 44.5 51.1 57.7 63.7
## 6 Transport and storage 18.2 21.1 23.6 21.0
## slovenia southern
## 1 23.8 24.7
## 2 55.2 66.2
## 3 48.3 46.0
## 4 75.0 79.8
## 5 51.9 43.8
## 6 15.1 14.8Columns. Intangible share of total investment, 2008-2021, by industry and peer group. Percent.
# Slovenia's distance from each peer group, in percentage points
peer_means <- sections %>%
inner_join(geo_lookup, by = "geo") %>%
filter(year >= 2008, peer_group != "slovenia") %>%
group_by(industry, peer_group) %>%
summarise(peer_share = mean(intang_share, na.rm = TRUE), .groups = "drop")
si_means <- si %>%
filter(year >= 2008) %>%
group_by(industry) %>%
summarise(si_share = mean(intang_share, na.rm = TRUE), .groups = "drop")
peer_means %>%
inner_join(si_means, by = "industry") %>%
mutate(gap_pp = round(si_share - peer_share, 1)) %>%
select(industry, peer_group, gap_pp) %>%
tidyr::pivot_wider(names_from = peer_group, values_from = gap_pp) %>%
as.data.frame()## industry cee continental non_eu nordic
## 1 Accommodation and food service 2.7 -8.3 -23.7 -23.8
## 2 Information and communication -4.7 -3.0 -13.1 -17.6
## 3 Manufacturing 13.8 -8.0 -17.6 -21.9
## 4 Professional, scientific and technical 5.9 -6.8 -8.4 -11.0
## 5 Trade and repair of motor vehicles 7.4 0.8 -5.8 -11.9
## 6 Transport and storage -3.1 -6.0 -8.6 -5.9
## southern
## 1 -0.9
## 2 -11.0
## 3 2.3
## 4 -4.9
## 5 8.1
## 6 0.3Columns. gap_pp = Slovenia’s intangible
share minus the peer group’s, in percentage points. Negative means
Slovenia is below that group.
12 Division level
Economic competencies only, since national-accounts investment does not exist below section level for Slovenia.
12.0.1 Excluding divisions too small to measure
A ratio computed on a very small denominator is unstable and can be meaningless. Coke and refined petroleum (C19) has an average adjusted value added of 0.7 million euro - one hundredth of one percent of Slovenian manufacturing, and roughly 400 times smaller than the next smallest division. Its apparent economic-competencies intensity of 15.8 percent, of which 7.5 percentage points is workforce training, is not credible: it is a rounding error in the denominator.
Rather than excluding it by name, the rule below drops any division worth less than 0.1 percent of its parent section, and reports each division’s size so the reader can judge. At present this catches C19 and nothing else.
division_size <- divisions %>%
filter(geo == "SI", year >= 2008) %>%
group_by(parent, nace) %>%
summarise(VAadj_mEUR = mean(VAadj, na.rm = TRUE), .groups = "drop") %>%
group_by(parent) %>%
mutate(pct_of_parent = round(100 * VAadj_mEUR / sum(VAadj_mEUR), 2)) %>%
ungroup()
division_size %>%
filter(pct_of_parent < 0.1) %>%
as.data.frame()## parent nace VAadj_mEUR pct_of_parent
## 1 C C19 1.011284 0.01Columns. VAadj_mEUR = mean adjusted
gross value added in millions of euro. pct_of_parent = the
division’s share of its section. Only divisions falling below the 0.1
percent threshold are listed - these are excluded below.
divisions %>%
filter(geo == "SI", year >= 2008) %>%
inner_join(division_size %>% select(nace, VAadj_mEUR, pct_of_parent),
by = "nace") %>%
filter(pct_of_parent >= 0.1) %>% # drop divisions too small to measure
mutate(ec = 100 * I_EconComp / VAadj,
brand = 100 * I_Brand / VAadj,
orgcap = 100 * I_OrgCap / VAadj,
training = 100 * I_Train / VAadj) %>%
group_by(parent, nace, industry, VAadj_mEUR, pct_of_parent) %>%
summarise(econ_competencies = round(mean(ec, na.rm = TRUE), 2),
brand = round(mean(brand, na.rm = TRUE), 2),
orgcap = round(mean(orgcap, na.rm = TRUE), 2),
training = round(mean(training, na.rm = TRUE), 2),
.groups = "drop") %>%
arrange(parent, desc(econ_competencies)) %>%
as.data.frame()## parent nace
## 1 C C21
## 2 C C10-C12
## 3 C C27
## 4 C C22-C23
## 5 C C26
## 6 C C20
## 7 C C24-C25
## 8 C C28
## 9 C C31-C33
## 10 C C13-C15
## 11 C C16-C18
## 12 C C29-C30
## 13 G G46
## 14 G G45
## 15 G G47
## 16 H H51
## 17 H H52
## 18 H H49
## 19 H H53
## 20 H H50
## 21 J J58-J60
## 22 J J62-J63
## 23 J J61
## industry
## 1 Manufacture of basic pharmaceutical products and pharmaceutical preparations
## 2 Manufacture of food products; beverages and tobacco products
## 3 Manufacture of electrical equipment
## 4 Manufacture of rubber and plastic products and other non-metallic mineral products
## 5 Manufacture of computer, electronic and optical products
## 6 Manufacture of chemicals and chemical products
## 7 Manufacture of basic metals and fabricated metal products, except machinery and equipment
## 8 Manufacture of machinery and equipment n.e.c.
## 9 Manufacture of furniture; jewellery, musical instruments, toys; repair and installation of machinery and equipment
## 10 Manufacture of textiles, wearing apparel, leather and related products
## 11 Manufacture of wood, paper, printing and reproduction
## 12 Manufacture of motor vehicles, trailers, semi-trailers and of other transport equipment
## 13 Wholesale trade, except of motor vehicles and motorcycles
## 14 Wholesale and retail trade and repair of motor vehicles and motorcycles
## 15 Retail trade, except of motor vehicles and motorcycles
## 16 Air transport
## 17 Warehousing and support activities for transportation
## 18 Land transport and transport via pipelines
## 19 Postal and courier activities
## 20 Water transport
## 21 Publishing, motion picture, video, television programme production; sound recording, programming and broadcasting activities
## 22 Computer programming, consultancy, and information service activities
## 23 Telecommunications
## VAadj_mEUR pct_of_parent econ_competencies brand orgcap training
## 1 1108.31580 12.71 9.98 7.26 1.38 1.34
## 2 612.70962 7.03 8.47 5.14 2.63 0.69
## 3 859.24496 9.85 6.10 1.91 2.52 1.67
## 4 940.84741 10.79 5.94 1.33 3.62 0.99
## 5 298.70247 3.43 5.90 1.21 3.11 1.59
## 6 409.57567 4.70 5.79 1.57 2.82 1.40
## 7 1588.03980 18.21 5.66 0.97 3.07 1.63
## 8 620.06648 7.11 5.43 1.19 3.19 1.05
## 9 666.07497 7.64 5.28 1.30 2.99 0.99
## 10 286.74215 3.29 5.03 1.28 3.07 0.69
## 11 629.67477 7.22 4.86 1.31 2.93 0.62
## 12 698.68652 8.01 4.73 1.12 2.57 1.03
## 13 2202.59041 46.57 9.46 4.93 3.61 0.92
## 14 586.59979 12.40 8.18 3.84 2.64 1.70
## 15 1940.88864 41.03 7.69 4.00 3.14 0.55
## 16 36.15627 1.61 6.58 1.61 3.17 1.80
## 17 771.72330 34.36 3.07 0.94 1.46 0.67
## 18 1193.46015 53.13 3.05 0.40 2.17 0.49
## 19 202.62900 9.02 2.80 0.73 1.74 0.33
## 20 42.18290 1.88 2.75 0.33 1.72 0.70
## 21 292.81496 18.34 12.81 5.59 6.22 1.00
## 22 745.86242 46.71 7.39 3.16 3.64 0.59
## 23 558.06302 34.95 6.43 3.79 2.47 0.17Columns. All percentages of adjusted gross value
added. econ_competencies is the total; brand,
orgcap and training are its three components
and sum to it. VAadj_mEUR and pct_of_parent
show how large the division is, so a ratio resting on a small
denominator is visible rather than hidden.
What to look for. Which component dominates differs sharply by division - pharmaceuticals and food are brand-led, most of manufacturing is organisational capital, air transport is training and organisational capital. Inside information and communication, publishing and media invests roughly twice what software and consultancy does, which the section-level figures conceal.
Columns. econ_competencies_pct_VA =
investment in brand, organisational capital and training ÷ adjusted
gross value added, percent. Divisions only — national-accounts
investment does not exist below section level for Slovenia, so total and
tangible investment cannot be shown here.
Within information and communication, publishing and media invests roughly twice what software and consultancy does in economic competencies - the reverse of what the section-level software story would suggest. Nikolov et al. (Mid-Tech Europe?) find that “most TFP growth can be attributed to a relatively few industries”; this suggests the same concentration may hold below section level, and that section-level aggregates conceal it. Estimated data only - no national-accounts anchor exists at this depth.
12.0.2 Figure
divisions %>%
filter(geo == "SI", year >= 2008) %>%
inner_join(division_size %>% select(nace, pct_of_parent), by = "nace") %>%
filter(pct_of_parent >= 0.1) %>%
mutate(ec = 100 * I_EconComp / VAadj) %>%
group_by(parent, nace) %>%
summarise(ec = mean(ec, na.rm = TRUE), .groups = "drop") %>%
ggplot(aes(stats::reorder(nace, ec), ec)) +
geom_col(width = 0.7, fill = "grey55", colour = "black", linewidth = 0.2) +
coord_flip() +
facet_wrap(~ parent, ncol = 2, scales = "free_y") +
labs(title = "Investment in economic competencies, Slovenian divisions",
subtitle = "brand, organisational capital and training, percent of adjusted gross value added, 2008-2021",
x = NULL, y = "percent") +
theme_pkp()
Figure 6. Investment in economic competencies by division, Slovenia, 2008-2021, percent of adjusted gross value added. Divisions below 0.1 percent of their section are excluded. Source: EUKLEMS & INTANProd, own calculations.