Adventures of a self-learning data scientist - Cross-country productivity analysis is orange to apple comparison

Code

library(tidyverse)
library(lubridate)
library(patchwork)

library(scales)
theme_set(theme_light())

library(OECD)
library(gghighlight)

Productivity comparison between the U.S. and Japan

Prof Tetsuji Okazaki of the University of Tokyo shows Japan lags the US in labor productivity level, and the gap is widening, in his “Lecture on Economics” in Nikkei news paper on February 17, 2023 (Japanese).

Code

countries <- c("USA", "JPN", "GBR", "DEU", "FRA", "ITA", "CAN")

exchange_rate <- get_dataset(
  "SNA_TABLE4",
  filter = list(countries, c("PPPGDP")),
  start_time = 2005, end_time = 2021
) |> 
  mutate(across(c(ObsValue, Time), as.numeric))

gdp <- get_dataset(
  "SNA_TABLE1",
  filter = list(countries, "B1_GE", c("DOB", "C", "V")),
  start_time = 2005, end_time = 2021
) |> 
  mutate(across(c(ObsValue, Time), as.numeric))

labor <- get_dataset(
  "ALFS_SUMTAB",
  filter = list(countries, "YGTT08L1_ST"),
  start_time = 2005, end_time = 2021
) |> 
  mutate(across(c(ObsValue, Time), as.numeric))

hours_worked <- get_dataset(
  "ANHRS",
  filter = list(countries, "TE"),
  start_time = 2005, end_time = 2021
) |> 
  mutate(across(c(ObsValue, Time), as.numeric))

Code

us_deflator <- gdp |> 
  filter(LOCATION == "USA", MEASURE == "DOB") |> 
  mutate(deflator = ObsValue / 100) |> 
  select(Time, deflator)

employment <- labor |> 
  select(LOCATION, Time, employment = ObsValue)

ppp <- exchange_rate |> 
  select(LOCATION, Time, ppp = ObsValue)

Code

ppp_method <- gdp |> 
  filter(MEASURE == "C") |> 
  left_join(employment, by = c("LOCATION", "Time")) |> 
  left_join(ppp, by = c("LOCATION", "Time")) |> 
  left_join(us_deflator, by = "Time") |> 
  mutate(
    nominal_prod_loc = ObsValue / employment,
    nominal_prod_usd = nominal_prod_loc / ppp,
    real_prod_usd = nominal_prod_usd / deflator
    ) |> 
  group_by(LOCATION) |> 
  mutate(index = real_prod_usd / real_prod_usd[1] * 100) |> 
  ungroup()
  
p1 <- ppp_method |> 
  filter(LOCATION %in% c("USA", "JPN")) |> 
  mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> 
  ggplot(aes(Time, real_prod_usd)) +
  geom_line(aes(color = LOCATION)) +
  labs(x = NULL, y = "Real productivity per person\n(thousand dollars, PPP, 2015 price)",
       color = "Country")

p2 <- ppp_method |> 
  filter(LOCATION %in% c("USA", "JPN")) |> 
  mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> 
  ggplot(aes(Time, index)) +
  geom_line(aes(color = LOCATION)) +
  scale_y_continuous(limits = c(85, 125)) +
  labs(x = NULL, y = "Real productivity per person\n(2005 = 100)",
       color = "Country")

Code

hours_worked_te <- hours_worked |> 
  select(COUNTRY, Time, hours_worked = ObsValue)

non_ppp_method <- gdp |> 
  filter(MEASURE == "V") |> 
  left_join(employment, by = c("LOCATION", "Time")) |> 
  left_join(hours_worked_te, by = c("LOCATION" = "COUNTRY", "Time")) |> 
  mutate(
    real_prod_loc = ObsValue / employment,
    real_prod_loc_hr = real_prod_loc / hours_worked
    ) |> 
  group_by(LOCATION) |> 
  mutate(
    prod_per_person = real_prod_loc / real_prod_loc[1] * 100,
    prod_per_hour = real_prod_loc_hr / real_prod_loc_hr[1] * 100
    ) |> 
  ungroup()

p3 <- non_ppp_method |> 
  filter(LOCATION %in% c("USA", "JPN")) |> 
  mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> 
  ggplot(aes(Time, prod_per_person)) +
  geom_line(aes(color = LOCATION)) +
  geom_line(aes(y = index), color = "gray80",
            data = ppp_method |> filter(LOCATION == "JPN")) +
  scale_y_continuous(limits = c(85, 125)) +
  labs(x = NULL, y = "Real productivity per person\n(2005 = 100)",
       color = "Country")

p4 <- non_ppp_method |> 
  filter(LOCATION %in% c("USA", "JPN")) |> 
  mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> 
  ggplot(aes(Time, prod_per_hour)) +
  geom_line(aes(color = LOCATION)) +
  scale_y_continuous(limits = c(85, 125)) +
  labs(x = NULL, y = "Real productivity per hour\n(2005 = 100)",
       color = "Country")

He shows the chart which is not exactly same, but similar to the left chart below, and says the gap in labor productivity level between the US and Japan is widening. I have modified it to make the right chart so that 2005 = 100. He uses PPP method. I think it is very difficult or impossible to compare the productivity level, as Japan and the US produce very different products and services. It is only possible to compare productivity growth, then using PPP is unnecessary and inappropriate.

Code

p1 + p2 + plot_layout(ncol = 2, guides = "collect") + plot_annotation(
  title = "PPP method: nominal GDP per employed person in JPY\nconverted to USD by PPP, then deflated to real by US GDP deflator",
  caption = "Source: OECD stat"
)

Figure 1: Per-person productivity in 2015 dollars (PPP) and its index (2005 = 100)

Below is the estimates of labor productivity growth without using PPP. Growth gap narrows in the left below. If we look at productivity per hour instead of per person in the right below, the gap narrows even further.

Code

p3 + p4 + plot_layout(ncol = 2, guides = "collect") + plot_annotation(
  title = "Non PPP method: real GDP per employed person or per hour in JPY \nindexed at 2005 = 100",
  subtitle = "Gray line in the left is JPN by PPP method, the same as above right",
  caption = "Source: OECD stat"
)

Figure 2: Per-person and per-hour productivity index (2005 = 100)

Productivity comparison among G7 countries

Although Japan looks like a loser in productivity per person in PPP dollars, it is not a loser in growth of productivity per hour.

Code

p5 <- ppp_method |> 
  ggplot(aes(Time, real_prod_usd)) +
  geom_line(aes(color = LOCATION)) +
  gghighlight(LOCATION == "JPN", use_group_by = FALSE) +
  labs(x = NULL, y = "Real productivity per person\n(thousand dollars, PPP, 2015 price)",
       color = "Country")

p6 <- non_ppp_method |> 
  ggplot(aes(Time, prod_per_person)) +
  geom_line(aes(color = LOCATION)) +
  gghighlight(LOCATION == "JPN", use_group_by = FALSE) +
  scale_y_continuous(limits = c(85, 125)) +
  labs(x = NULL, y = "Real productivity per person (2005 = 100)",
       color = "Country")

p7 <- non_ppp_method |> 
  ggplot(aes(Time, prod_per_hour)) +
  geom_line(aes(color = LOCATION)) +
  gghighlight(LOCATION == "JPN", use_group_by = FALSE) +
  scale_y_continuous(limits = c(85, 125)) +
  labs(x = NULL, y = "Real productivity per hour (2005 = 100)",
       color = "Country")

p5 + p6 + p7 + plot_layout(ncol = 3, guides = "collect") + plot_annotation(
  title = "Japan is not a loser in growth of productivity per hour\namong G7 countries",
  caption = "Source: OECD stat"
)

Figure 3: Per-person productivity in 2015 dollars (PPP), per-person and per-hour productivity index (2005 = 100)

--- title: Cross-country productivity analysis is orange to apple comparison author: Mitsuo Shiota date: '2023-02-17' date-modified: '2023-06-14' categories: - economics - Nikkei knitr: opts_chunk: out.width: '70%' --- ```{r} #| label: setup #| message: false library(tidyverse) library(lubridate) library(patchwork) library(scales) theme_set(theme_light()) library(OECD) library(gghighlight) ``` ## Productivity comparison between the U.S. and Japan [Prof Tetsuji Okazaki of the University of Tokyo](https://www.e.u-tokyo.ac.jp/fservice/faculty/okazaki/okazaki-e/okazaki01-e.html) shows Japan lags the US in labor productivity level, and the gap is widening, in his ["Lecture on Economics" in Nikkei news paper on February 17, 2023 (Japanese)](https://www.nikkei.com/article/DGXZQOCD2515M0V20C23A1000000/). ```{r} #| label: get_data #| message: false countries <- c("USA", "JPN", "GBR", "DEU", "FRA", "ITA", "CAN") exchange_rate <- get_dataset( "SNA_TABLE4", filter = list(countries, c("PPPGDP")), start_time = 2005, end_time = 2021 ) |> mutate(across(c(ObsValue, Time), as.numeric)) gdp <- get_dataset( "SNA_TABLE1", filter = list(countries, "B1_GE", c("DOB", "C", "V")), start_time = 2005, end_time = 2021 ) |> mutate(across(c(ObsValue, Time), as.numeric)) labor <- get_dataset( "ALFS_SUMTAB", filter = list(countries, "YGTT08L1_ST"), start_time = 2005, end_time = 2021 ) |> mutate(across(c(ObsValue, Time), as.numeric)) hours_worked <- get_dataset( "ANHRS", filter = list(countries, "TE"), start_time = 2005, end_time = 2021 ) |> mutate(across(c(ObsValue, Time), as.numeric)) ``` ```{r} #| label: us_deflator us_deflator <- gdp |> filter(LOCATION == "USA", MEASURE == "DOB") |> mutate(deflator = ObsValue / 100) |> select(Time, deflator) employment <- labor |> select(LOCATION, Time, employment = ObsValue) ppp <- exchange_rate |> select(LOCATION, Time, ppp = ObsValue) ``` ```{r} #| label: ppp_method ppp_method <- gdp |> filter(MEASURE == "C") |> left_join(employment, by = c("LOCATION", "Time")) |> left_join(ppp, by = c("LOCATION", "Time")) |> left_join(us_deflator, by = "Time") |> mutate( nominal_prod_loc = ObsValue / employment, nominal_prod_usd = nominal_prod_loc / ppp, real_prod_usd = nominal_prod_usd / deflator ) |> group_by(LOCATION) |> mutate(index = real_prod_usd / real_prod_usd[1] * 100) |> ungroup() p1 <- ppp_method |> filter(LOCATION %in% c("USA", "JPN")) |> mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> ggplot(aes(Time, real_prod_usd)) + geom_line(aes(color = LOCATION)) + labs(x = NULL, y = "Real productivity per person\n(thousand dollars, PPP, 2015 price)", color = "Country") p2 <- ppp_method |> filter(LOCATION %in% c("USA", "JPN")) |> mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> ggplot(aes(Time, index)) + geom_line(aes(color = LOCATION)) + scale_y_continuous(limits = c(85, 125)) + labs(x = NULL, y = "Real productivity per person\n(2005 = 100)", color = "Country") ``` ```{r} #| label: non_ppp_method hours_worked_te <- hours_worked |> select(COUNTRY, Time, hours_worked = ObsValue) non_ppp_method <- gdp |> filter(MEASURE == "V") |> left_join(employment, by = c("LOCATION", "Time")) |> left_join(hours_worked_te, by = c("LOCATION" = "COUNTRY", "Time")) |> mutate( real_prod_loc = ObsValue / employment, real_prod_loc_hr = real_prod_loc / hours_worked ) |> group_by(LOCATION) |> mutate( prod_per_person = real_prod_loc / real_prod_loc[1] * 100, prod_per_hour = real_prod_loc_hr / real_prod_loc_hr[1] * 100 ) |> ungroup() p3 <- non_ppp_method |> filter(LOCATION %in% c("USA", "JPN")) |> mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> ggplot(aes(Time, prod_per_person)) + geom_line(aes(color = LOCATION)) + geom_line(aes(y = index), color = "gray80", data = ppp_method |> filter(LOCATION == "JPN")) + scale_y_continuous(limits = c(85, 125)) + labs(x = NULL, y = "Real productivity per person\n(2005 = 100)", color = "Country") p4 <- non_ppp_method |> filter(LOCATION %in% c("USA", "JPN")) |> mutate(LOCATION = factor(LOCATION, levels = c("USA", "JPN"))) |> ggplot(aes(Time, prod_per_hour)) + geom_line(aes(color = LOCATION)) + scale_y_continuous(limits = c(85, 125)) + labs(x = NULL, y = "Real productivity per hour\n(2005 = 100)", color = "Country") ``` He shows the chart which is not exactly same, but similar to the left chart below, and says the gap in labor productivity level between the US and Japan is widening. I have modified it to make the right chart so that 2005 = 100. He uses PPP method. I think it is very difficult or impossible to compare the productivity level, as Japan and the US produce very different products and services. It is only possible to compare productivity growth, then using PPP is unnecessary and inappropriate. ```{r} #| label: fig-ppp_method_chart #| fig-cap: Per-person productivity in 2015 dollars (PPP) and its index (2005 = 100) #| fig-align: center p1 + p2 + plot_layout(ncol = 2, guides = "collect") + plot_annotation( title = "PPP method: nominal GDP per employed person in JPY\nconverted to USD by PPP, then deflated to real by US GDP deflator", caption = "Source: OECD stat" ) ``` Below is the estimates of labor productivity growth without using PPP. Growth gap narrows in the left below. If we look at productivity per hour instead of per person in the right below, the gap narrows even further. ```{r} #| label: fig-non_ppp_method_chart #| fig-cap: Per-person and per-hour productivity index (2005 = 100) #| fig-align: center p3 + p4 + plot_layout(ncol = 2, guides = "collect") + plot_annotation( title = "Non PPP method: real GDP per employed person or per hour in JPY \nindexed at 2005 = 100", subtitle = "Gray line in the left is JPN by PPP method, the same as above right", caption = "Source: OECD stat" ) ``` ## Productivity comparison among G7 countries Although Japan looks like a loser in productivity per person in PPP dollars, it is not a loser in growth of productivity per hour. ```{r} #| label: fig-g7 #| message: false #| warning: false #| fig-cap: Per-person productivity in 2015 dollars (PPP), per-person and per-hour productivity index (2005 = 100) #| fig-align: center p5 <- ppp_method |> ggplot(aes(Time, real_prod_usd)) + geom_line(aes(color = LOCATION)) + gghighlight(LOCATION == "JPN", use_group_by = FALSE) + labs(x = NULL, y = "Real productivity per person\n(thousand dollars, PPP, 2015 price)", color = "Country") p6 <- non_ppp_method |> ggplot(aes(Time, prod_per_person)) + geom_line(aes(color = LOCATION)) + gghighlight(LOCATION == "JPN", use_group_by = FALSE) + scale_y_continuous(limits = c(85, 125)) + labs(x = NULL, y = "Real productivity per person (2005 = 100)", color = "Country") p7 <- non_ppp_method |> ggplot(aes(Time, prod_per_hour)) + geom_line(aes(color = LOCATION)) + gghighlight(LOCATION == "JPN", use_group_by = FALSE) + scale_y_continuous(limits = c(85, 125)) + labs(x = NULL, y = "Real productivity per hour (2005 = 100)", color = "Country") p5 + p6 + p7 + plot_layout(ncol = 3, guides = "collect") + plot_annotation( title = "Japan is not a loser in growth of productivity per hour\namong G7 countries", caption = "Source: OECD stat" ) ```