source(here::here("scripts/load-libraries.R"))
theme_set(theme_light() +
theme(panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
plot.title = element_text(size = rel(2))))
my_caption_rbf = "Daniel Moul. Source: Reserve Bank of Fiji"
my_caption_fbos = "Daniel Moul. Source: Fiji Bureau of Statistics via Reserve Bank of Fiji"The data used in this section comes from the Reserve Bank of Fiji and the US Federal Reserve Bank of Minneaopolis.
The Reserve Bank of Fiji makes available historical consumer price data going back to 1991, rebased in 2005, 2008, 2011, and 2014. The CPI is based on the changing prices of a basket of representative products and services grouped in CPI component categories, and then the components are weighted to reflect consumer spending. The component categories changed between 2005 and 2008 (making comparisons between them difficult). Fortunately, only the weightings changed in subsequent rebases. Where year-to-year price changes were included in multiple rebases, I averaged them to calculate a single, simple, non-authoritative CPI for the entire period (Figure 9.1 panel B).
cpi_2005_base_raw <- read_xlsx(here("data/raw/rbf/7.1-Consumer-Price-Index-fixed.xlsx"),
range = "A3:U28") |>
clean_names() |>
remove_empty(which = c("rows", "cols"))
cpi_2008_base_raw <- read_xlsx(here("data/raw/rbf/7.1-Consumer-Price-Index-fixed.xlsx"),
range = "A33:U40") |>
clean_names() |>
remove_empty(which = c("rows", "cols"))
cpi_2011_base_raw <- read_xlsx(here("data/raw/rbf/7.1-Consumer-Price-Index-fixed.xlsx"),
range = "A45:U55") |>
clean_names() |>
remove_empty(which = c("rows", "cols"))
cpi_2014_base_raw <- read_xlsx(here("data/raw/rbf/7.1-Consumer-Price-Index-fixed.xlsx"),
range = "A60:U70") |>
clean_names() |>
remove_empty(which = c("rows", "cols"))
cpi_merged <- bind_rows(
cpi_2005_base_raw |>
mutate(base = 2005),
cpi_2008_base_raw |>
mutate(base = 2008),
cpi_2011_base_raw |>
mutate(base = 2011),
cpi_2014_base_raw |>
mutate(base = 2014)
) |>
rename(communications = communi_cations)
cpi_merged_long <- cpi_merged |>
filter(base != 2005,
period != "Weight") |>
remove_empty(which = "cols") |>
select(-year_on_year_percent_change, -all_items) |>
relocate(base, .before = period) |>
mutate(period = as.numeric(period),
base = glue("base: {base}")) |>
pivot_longer(
cols = clothing_footwear:recreation_culture,
names_to = "metric",
values_to = "value"
) |>
mutate(metric = case_match(
metric,
"clothing_footwear" ~ "clothing_footwear",
"transport" ~ "transport",
"food_and_non_alcoholic_beverage" ~ "food_non_alcohol_bev",
"alcoholic_beverages_tobacco_and_narcotics" ~ "alcohol_tobacco_etc",
"housing_water_elec_gas_and_other_fuels" ~ "housing_water_elec_etc",
"furnishings_hhld_equip_routine_hhld_maint" ~ "furnishings_hhld_etc",
"health" ~ "health",
"communications" ~ "communications",
"recreation_culture" ~ "recreation_culture",
.default = metric # default just in case; shouldn't be needed
))
cpi_weights <- cpi_merged |>
filter(base != 2005) |>
remove_empty(which = "cols")
cpi <- cpi_merged |>
select(period, year_on_year_percent_change, base) |>
filter(period != "Weight") |>
mutate(year = as.numeric(period)) |>
summarize(yty = mean(year_on_year_percent_change) / 100,
cpi = 0,
.by = year) #|>
# mutate(cpi = lag(cpi) * (1 + yty)) #lag(yty, default = 0)))
# since lag() doesn't seem to get the whole vector at the beginning of the call, I cannot
# modify cpi while lagging it. So instead I use the following:
for(i in 1:length(cpi$yty)) {
if(i == 1) {
cpi$cpi[i] <- 1
} else {
cpi$cpi[i] <- cpi$cpi[i - 1] * (1 + cpi$yty[i])
}
}
cpi_year_min <- min(cpi$year)
cpi_year_max <- max(cpi$year)
cpi_weights_tmp <- cpi_merged |>
filter(base != 2005) |>
remove_empty(which = "cols") |>
relocate(base) |>
select(-all_items) |>
pivot_longer(
cols = clothing_footwear:recreation_culture,
names_to = "metric",
values_to = "weight"
) |>
mutate(metric = case_match(
metric,
"clothing_footwear" ~ "clothing_footwear",
"transport" ~ "transport",
"food_and_non_alcoholic_beverage" ~ "food_non_alcohol_bev",
"alcoholic_beverages_tobacco_and_narcotics" ~ "alcohol_tobacco_etc",
"housing_water_elec_gas_and_other_fuels" ~ "housing_water_elec_etc",
"furnishings_hhld_equip_routine_hhld_maint" ~ "furnishings_hhld_etc",
"health" ~ "health",
"communications" ~ "communications",
"recreation_culture" ~ "recreation_culture"
))
cpi_weights <- cpi_weights_tmp |>
filter(base != 2005,
period == "Weight") |>
remove_empty(which = "cols") |>
relocate(base) |>
select(-period)
dta_for_cpi_weight_plot <- cpi_weights_tmp |>
filter(base != 2005,
period != "Weight") |>
remove_empty(which = "cols") |>
relocate(base) |>
rename(value = weight) |>
left_join(
cpi_weights,
by = c("base", "metric")
)
us_cpi <- read_xlsx(here("data/raw/us-frb/us-cpi-historical.xlsx"),
skip = 7) |>
clean_names() |>
mutate(cpi_base_1991 = cpi / cpi[year == 1991])p1 <- cpi |>
ggplot(aes(year, yty)) +
geom_hline(yintercept = 0, lty = 2, linewidth = 0.15, alpa = 0.5) +
geom_line() +
geom_line(data = us_cpi |>
filter(year >= 1991),
aes(year, annual_pct_change),
color = "firebrick", linewidth = 0.2,
alpha = 0.75) +
annotate("text", x = 2002, y = -0.01, label = "US yty pct change",
hjust = 0, color = "firebrick", alpha = 0.75) +
scale_y_continuous(labels = label_percent()) +
labs(
subtitle = glue("A: Year to year change in consumer prices",
"\nIn Fiji and US")
)
p2 <- cpi |>
ggplot() +
geom_line(aes(year, cpi)) +
geom_line(data = us_cpi |>
filter(year >= 1991),
aes(year, cpi_base_1991),
color = "firebrick", linewidth = 0.2,
alpha = 0.75) +
annotate("text", x = 2011, y = 1.5, label = "US CPI",
hjust = 0, color = "firebrick", alpha = 0.75) +
labs(
subtitle = glue("B: Fiji prices doubled in ~20 years",
"\nwhile US prices doubled in ~30 years",
"\nCPI: {cpi_year_min} = 1.0"),
y = "Consumer price index"
)
p1 + p2 +
plot_annotation(
title = "Consumer price changes in Fiji",
subtitle = glue("{cpi_year_min} to {cpi_year_max}. With US comparison."),
caption = paste0(my_caption_rbf,
" and Minneapolis Federal Reserve Bank")
)
“Food and non-alcoholic beverages” is heaviest weighted component of the CPI, and price increases in this component since the COVID-19 pandemic have been especially painful. While I was in Fiji friends and relatives were talking about the current “affordability crisis”.
In Figure 9.2 and Figure 9.3 \[component\ weighted\ value = component\ weight * relative \ price\] and scaled such that the \[\sum{component\ weighted\ values} = 1.0,\ for\ all\ component\ weighted\ values\ in\ 2023\]
dta_for_plot <- dta_for_cpi_weight_plot |>
mutate(base = glue("base: {base}"),
period = as.numeric(period),
weighted_value = weight * value,
metric = fct_reorder(metric, -weighted_value, min),
weighted_value_scaled = weighted_value / sum(weighted_value[period == 2023])
)
dta_for_plot |>
ggplot(aes(period, weighted_value_scaled, color = base, group = base)) +
geom_line() +
scale_x_continuous(breaks = c(2010, 2015, 2020, 2023)) +
facet_wrap(~metric) +
guides(color = guide_legend(ncol = 3,
override.aes = list(linewidth = 3))
) +
theme(legend.position = "bottom") +
labs(
title = glue("The most signficant changes for consumers were in the price of",
"\nfood and non-alcoholic beverages"),
subtitle = glue("Relative weight of the components in the Fiji CPI and changes in their prices."),
x = NULL,
y = "Weighted value",
color = NULL,
caption = my_caption_rbf
)
Figure 9.3 presents the same data with varying Y axes, which makes the trends within components more visible.
dta_for_plot |>
ggplot(aes(period, weighted_value_scaled, color = base, group = base)) +
geom_line() +
scale_x_continuous(breaks = c(2010, 2015, 2020, 2023)) +
facet_wrap(~metric, scales = "free_y") +
guides(color = guide_legend(ncol = 3,
override.aes = list(linewidth = 3))
) +
theme(legend.position = "bottom") +
labs(
title = glue("The most signficant changes for consumers were in the price of",
"\nfood and non-alcoholic beverages (Y axis scale varies)"),
subtitle = glue("Changes in prices of the components of Fiji CPI"),
x = NULL,
y = "Weighted value",
color = NULL,
caption = my_caption_rbf
)
The components and their weightings reflect the Reserve Bank’s understanding of how people in Fiji spend their money. In each rebase, the reweightings were scaled to sum to 1000 (Figure 9.4 panel A).
p1 <- dta_for_cpi_weight_plot |>
mutate(base = glue("base: {base}"),
period = as.numeric(period)) |>
ggplot(aes(period, weight, color = base, group = base)) +
geom_hline(yintercept = 0, lty = 2, linewidth = 0.15, alpa = 0.5) +
geom_line() +
guides(color = guide_legend(override.aes = list(linewidth = 3))) +
facet_wrap(~metric) +
theme(legend.position = "bottom") +
labs(
subtitle = "A: Rebased components",
x = NULL,
y = "Relative weighting",
color = NULL
)
p2 <- dta_for_cpi_weight_plot |>
mutate(base = glue("base: {base}"),
period = as.numeric(period)) |>
ggplot(aes(period, weight, fill = metric)) +
geom_col(position = position_fill(),
alpha = 0.6,
na.rm = TRUE) +
scale_x_continuous(expand = expansion(mult = c(0,0)),
breaks = 2010 + 2 * 0:10) +
scale_y_continuous(expand = expansion(mult = c(0,0))) +
theme(legend.position = "bottom") +
facet_wrap(~base, scales ="free_x") +
labs(
subtitle = "B: Components' relative contributions",
x = NULL,
y = NULL,
fill = NULL
)
p1 + p2 +
plot_annotation(
title = glue("Changes have been modest in the relative re-weightings",
"\nof Fiji CPI components"),
caption = my_caption_rbf
) +
guides(fill = guide_legend(ncol = 2))
Ignoring component weightings, Figure 9.5 and Figure 9.6 present the degree of inflation within each component.
cpi_merged_long |>
ggplot(aes(period, value, color = base, group = base)) +
geom_hline(yintercept = 100, lty = 2, linewidth = 0.15, alpa = 0.5) +
geom_line() +
scale_x_continuous(breaks = c(2010, 2015, 2020, 2023)) +
facet_wrap(~metric) +
guides(color = guide_legend(ncol = 3,
override.aes = list(linewidth = 3))
) +
theme(legend.position = "bottom") +
labs(
title = "Changes in prices of the components of Fiji CPI",
subtitle = "Year of base = 100",
x = NULL,
y = NULL,
color = NULL,
caption = my_caption_rbf
)
Figure 9.6 presents the same data with varying Y axes, which makes the trends within components more visible.
cpi_merged_long |>
ggplot(aes(period, value, color = base, group = base)) +
geom_hline(yintercept = 100, lty = 2, linewidth = 0.15, alpa = 0.5) +
geom_line() +
scale_x_continuous(breaks = c(2010, 2015, 2020, 2023)) +
facet_wrap(~metric, scales = "free_y") +
guides(color = guide_legend(ncol = 3,
override.aes = list(linewidth = 3))
) +
theme(legend.position = "bottom") +
labs(
title = "Changes in prices of the components of Fiji CPI (Y axis scale varies)",
subtitle = "Year of base = 100",
x = NULL,
y = NULL,
color = NULL,
caption = my_caption_rbf
)
Plotting all components for each rebase (Figure 9.7) makes visible the wide swings in relative prices in the ‘alcohol, tobacco and narcotics’ component.
cpi_merged_long |>
ggplot(aes(period, value, color = metric, group = metric)) +
geom_hline(yintercept = 100, lty = 2, linewidth = 0.15, alpa = 0.5) +
geom_line() +
scale_x_continuous(breaks = c(2010, 2015, 2020, 2023)) +
facet_wrap(~base) +
guides(color = guide_legend(ncol = 3,
override.aes = list(linewidth = 3))
) +
theme(legend.position = "bottom") +
labs(
title = "The largest price changes were in alcohol, tobacco and narcotics",
subtitle = glue("Changes in prices of the components of Fiji CPI",
"\nYear of base = 100"),
x = NULL,
y = NULL,
color = NULL,
caption = my_caption_rbf
)