Income, Output, and Complexity Indicators

What GDP and GNI per capita actually measure, why the real/nominal and PPP adjustments matter, and how the Economic Complexity Index captures productive capabilities that income measures miss.

1 Introduction

This page covers the indicators that describe what an economy produces and what its residents earn: GDP per capita, GNI per capita, and — as a measure of the productive capabilities behind those flows — the Economic Complexity Index.

These belong together because they answer versions of the same question. GDP and GNI ask how large the flow of output and income is. Economic complexity asks what kind of productive structure generates that flow. Neither is a measure of wellbeing, and the most common error in applied work is to treat them as if they were. The companion page on The Theory-Ladenness of Observation develops that point in general terms; here we look at what each measure does and does not contain.

library(WDI)
library(dplyr)
library(tidyr)
library(readr)
library(ggplot2)
library(ggrepel)
library(ggpubr)
library(kableExtra)

wdi_cached <- function(indicators, file, ...) {
  path <- file.path("data", file)
  if (!file.exists(path)) {
    message("Cache miss - downloading from the World Bank API ...")
    write_csv(WDI(indicator = indicators, ...), path)
  }
  out <- read_csv(path, show_col_types = FALSE)
  # Guard against a retired indicator code (which returns silently without its
  # column) and against a cache file written by a different query
  stopifnot(all(names(indicators) %in% names(out)))
  out
}

The wdi_cached() helper and the reasoning behind it are explained in Accessing Development Data in R.

2 Gross Domestic Product

2.1 Definition

GDP is “the market value of all final goods and services produced within a country in a given period” (Mankiw 2024, chap. 24).¹ It aggregates cars, haircuts, and smartphones by valuing each at its market price.

That aggregation rule has immediate consequences:

• Goods with no market price are largely excluded — subsistence production, household and care work, volunteering. Where they are included at all, it is through imputed estimates. This matters most in exactly the countries where such activity is largest.
• Only final demand counts. Steel sold to a car manufacturer is not added on top of the car.
• Only newly produced goods count. A new car enters GDP; the resale of a used one does not.

2.2 Two necessary adjustments

Real vs. nominal. Nominal GDP values output at each year’s own prices, so it rises when prices rise even if nothing more is produced. Real GDP holds prices fixed at a base year. If nominal GDP doubles while all prices double, real GDP is unchanged — and real GDP is the one that answers the question you almost always mean to ask.

Purchasing power parity. Market exchange rates do not reflect what money actually buys. 1,000 USD buys considerably more in India than in Norway, because the Indian price level is lower. PPP conversion re-values each country’s output at a common set of prices, expressed in “international dollars”. For comparisons of living standards across countries, PPP figures are the right choice; for questions about a country’s weight in the world economy or its capacity to import, market exchange rates are more appropriate.

NoteChoosing the right series

Question	Series	WDI code
Living standards across countries	GDP pc, PPP	NY.GDP.PCAP.PP.CD
Growth within one country over time	GDP pc, constant $	NY.GDP.PCAP.KD
Size in the world economy	GDP, current US$	NY.GDP.MKTP.CD

The Penn World Table is the standard alternative for long-run comparative work, distinguishing expenditure-side (rgdpe) from output-side (rgdpo) real GDP.

2.3 What GDP captures and misses

Captures: the market value of final output; standardised across countries through the UN System of National Accounts; long time series, back to 1960 for most countries; frequent updates; and a robust empirical correlation with many more fundamental welfare measures.

Misses: GDP was never designed as a welfare measure. It is silent on distribution, unpriced environmental damage and resource depletion, the informal economy and unpaid work, and the question of whether market prices track social value at all. Its correlation with welfare measures is not a licence to treat it as one.

3 Gross National Income

GNI equals GDP plus net primary income from abroad: income earned by residents abroad, minus income earned by non-residents domestically. GDP is about where production happens; GNI is about who receives the income.

The World Bank’s Atlas method smooths exchange-rate fluctuations using a three-year average, and GNI per capita computed this way is the basis for the familiar low-, lower-middle-, upper-middle- and high-income classifications.

3.1 Application: where GDP and GNI diverge

For most countries the two are close, and the divergences are precisely the interesting cases.

gdp_gni <- wdi_cached(
  indicators = c(
    GDPpc = "NY.GDP.PCAP.PP.CD",
    GNIpc = "NY.GNP.PCAP.PP.CD"
  ),
  file    = "wdi-gdp-gni.csv",
  country = c("US", "DE", "QA", "NO", "PH", "RO", "IE", "LU"),
  start   = 2000,
  end     = 2023
)

gdp_gni_avg <- gdp_gni |>
  filter(!is.na(GDPpc), !is.na(GNIpc)) |>
  summarise(across(c(GDPpc, GNIpc), mean), .by = country) |>
  mutate(
    difference     = GNIpc - GDPpc,
    difference_pct = 100 * difference / GDPpc
  )

ggplot(gdp_gni_avg, aes(x = GDPpc, y = GNIpc, colour = country)) +
  geom_abline(slope = 1, intercept = 0, linetype = "dashed", colour = "grey50") +
  geom_point(alpha = 0.8, size = 2) +
  geom_label_repel(aes(label = country), show.legend = FALSE, size = 3) +
  scale_colour_viridis_d(option = "H") +
  scale_x_continuous(labels = scales::dollar_format()) +
  scale_y_continuous(labels = scales::dollar_format()) +
  labs(
    x = "GDP per capita (PPP, current international $)",
    y = "GNI per capita (PPP, current international $)",
    caption = "Dashed line: GDP = GNI"
  ) +
  theme_minimal() +
  theme(legend.position = "none")

Figure 1: GDP and GNI per capita (PPP), averaged over 2000–2023. Points above the dashed line are countries whose residents receive more income than is produced domestically.

gdp_gni_avg |>
  select(country, difference_pct) |>
  arrange(desc(abs(difference_pct))) |>
  rename(Country = country, `GNI − GDP (%)` = difference_pct) |>
  kable(digits = 1) |>
  kable_styling(full_width = FALSE)

Table 1: Percentage difference between GNI and GDP per capita. Positive values mean GNI exceeds GDP.

Country	GNI − GDP (%)
Luxembourg	-26.9
Ireland	-21.3
Philippines	9.6
Qatar	-5.2
Germany	2.1
Romania	-1.7
United States	0.9
Norway	-0.5

The structural stories behind the gaps:

• Ireland and Luxembourg: GNI falls well short of GDP. Both host large numbers of foreign-owned multinationals whose profits are booked domestically and then repatriated. Irish GDP in particular is widely regarded as unusable as a welfare indicator — Ireland’s own statistical office publishes a modified GNI (GNI*) specifically to strip out these distortions.
• Qatar: GDP exceeds GNI because of income outflows to a large foreign workforce and to foreign-owned firms in the hydrocarbon sector.
• The Philippines: GNI exceeds GDP, driven by remittances from a very large overseas workforce.

The lesson is methodological rather than descriptive. GDP and GNI are close substitutes for most purposes, but the assumption that they are interchangeable fails exactly where a country’s integration into the world economy is unusual — which is often the case you are studying.

4 Economic Complexity

4.1 The idea

Economic complexity shifts attention from how much an economy produces to what it is able to produce. The underlying claim is that development consists in accumulating productive capabilities — the know-how, institutions, and infrastructure needed to make sophisticated things — and that what a country exports reveals which capabilities it has.²

The Economic Complexity Index (ECI) operationalises this by looking jointly at how diversified a country’s export basket is and how ubiquitous its products are. A country exporting many products that few others can make scores high; a country exporting a few products that everyone makes scores low.

A closely related construct is the product space: a network in which products are linked when they require similar capabilities. Peripheral products need few capabilities; core products need many, and only complex economies make them. The structure of the network implies that diversification is path-dependent — countries move most easily to products adjacent to what they already make.³

Sources: the Atlas of Economic Complexity (Harvard Growth Lab) and the Observatory of Economic Complexity.

Captures: productive capabilities invisible to GDP; a strong predictor of subsequent growth; concrete guidance on diversification opportunities; broad country and time coverage.

Misses: it is computed from exported goods, so services are poorly represented and non-traded activity is invisible; it requires high-quality trade data and so misses informal trade; the methodology is opaque to non-specialists; and it arguably embeds a bias toward manufacturing as the route to development.

4.2 Application: complexity, income, and the resource curse

Complexity correlates strongly with income. But there is an obvious alternative route to high income — exporting valuable natural resources — which requires no complexity at all. A good test of the complexity argument therefore has to ask how much of the relationship survives once resource wealth is accounted for.

eci <- read_csv("data/eci-rankings.csv", show_col_types = FALSE) |>
  filter(year == 2021) |>
  select(iso3c = country_iso3_code, ECI = eci_hs12)

resources <- wdi_cached(
  indicators = c(
    GDPpc = "NY.GDP.PCAP.PP.CD",
    Rents = "NY.GDP.TOTL.RT.ZS"   # total natural resource rents, % of GDP
  ),
  file    = "wdi-resource-rents.csv",
  country = "all",
  start   = 2021,
  end     = 2021,
  extra   = TRUE
)

complexity <- resources |>
  filter(!is.na(region), region != "Aggregates") |>
  select(country, iso3c, GDPpc, Rents) |>
  inner_join(eci, by = "iso3c") |>
  filter(if_all(c(ECI, GDPpc, Rents), \(x) !is.na(x)))

nrow(complexity)

[1] 139

r2 <- function(model) round(summary(model)$r.squared, 3)

m1 <- lm(log(GDPpc) ~ ECI, data = complexity)
m2 <- lm(log(GDPpc) ~ Rents, data = complexity)

complexity <- complexity |>
  mutate(
    eci_resid = residuals(lm(ECI ~ Rents, data = complexity)),
    gdp_resid = residuals(lm(log(GDPpc) ~ Rents, data = complexity))
  )
m3 <- lm(gdp_resid ~ eci_resid, data = complexity)

base_theme <- theme_minimal() +
  theme(plot.title = element_text(size = 11, face = "bold"),
        plot.subtitle = element_text(size = 9))

p1 <- ggplot(complexity, aes(x = ECI, y = GDPpc)) +
  geom_point(alpha = 0.7, colour = "#1f77b4", size = 2) +
  geom_smooth(method = "lm", se = TRUE, colour = "#d62728", linewidth = 1) +
  scale_y_log10(labels = scales::dollar_format(scale = 0.001, suffix = "k",
                                               accuracy = 1)) +
  labs(title = "Complexity and income", subtitle = paste0("R² = ", r2(m1)),
       x = "Economic Complexity Index", y = "GDP per capita (PPP, log)") +
  base_theme

p2 <- ggplot(complexity, aes(x = Rents, y = GDPpc)) +
  geom_point(alpha = 0.7, colour = "#ff7f0e", size = 2) +
  geom_smooth(method = "lm", se = TRUE, colour = "#d62728", linewidth = 1) +
  scale_y_log10(labels = scales::dollar_format(scale = 0.001, suffix = "k",
                                               accuracy = 1)) +
  labs(title = "Resource rents and income", subtitle = paste0("R² = ", r2(m2)),
       x = "Natural resource rents (% of GDP)", y = NULL) +
  base_theme

p3 <- ggplot(complexity, aes(x = eci_resid, y = gdp_resid)) +
  geom_point(alpha = 0.7, colour = "#2ca02c", size = 2) +
  geom_smooth(method = "lm", se = TRUE, colour = "#d62728", linewidth = 1) +
  labs(title = "Complexity and income,\ncontrolling for resources",
       subtitle = paste0("R² = ", r2(m3)),
       x = "ECI (residuals)", y = "log GDP per capita (residuals)") +
  base_theme

annotate_figure(
  ggarrange(p1, p2, p3, ncol = 3),
  bottom = text_grob(
    "Data: World Bank (GDP, resource rents, 2021); Economic Complexity Index, Harvard Growth Lab (HS12, 2021).",
    size = 9, hjust = 1, x = 1)
)

Figure 2: Left: complexity and income. Centre: resource rents and income. Right: the complexity–income relationship after partialling out resource rents from both variables.

Reading the three panels together: complexity explains a substantial share of the cross-country variation in income; resource rents on their own explain very little; and the complexity–income relationship is if anything slightly stronger once resources are partialled out. Complexity is not a proxy for resource wealth.

The countries sitting above the line in the left panel are richer than their capability base would predict, and these are overwhelmingly resource exporters:

complexity |>
  mutate(residual = residuals(m1)) |>
  slice_max(residual, n = 10) |>
  select(Country = country, ECI,
         `GDP pc (PPP)` = GDPpc, `Resource rents (% GDP)` = Rents) |>
  kable(digits = c(0, 2, 0, 1)) |>
  kable_styling(full_width = FALSE)

Table 2: The ten countries whose income most exceeds what their complexity predicts, ranked by the residual from the left-hand panel.

Country	ECI	GDP pc (PPP)	Resource rents (% GDP)
Qatar	-0.54	116833	27.3
Australia	-0.62	58327	13.4
Equatorial Guinea	-1.59	16821	23.5
Norway	0.60	92206	10.0
Oman	-0.39	38719	29.2
Kazakhstan	-0.54	32946	26.8
United Arab Emirates	0.37	68580	17.6
Mongolia	-1.30	14792	33.1
Chile	-0.49	29107	16.9
New Zealand	0.16	48202	1.5

complexity |>
  mutate(above_line = residuals(m1) >= sort(residuals(m1), decreasing = TRUE)[10]) |>
  summarise(mean_rents = mean(Rents), .by = above_line)

# A tibble: 2 × 2
  above_line mean_rents
  <lgl>           <dbl>
1 FALSE            6.80
2 TRUE            19.9 

Their average resource rents are roughly two and a half times the sample average. For these countries the “resource curse” literature suggests the position is more fragile than the income figure alone implies: income without capabilities is difficult to sustain once the resource is depleted or its price falls.

WarningWhat this analysis does not show

These are cross-sectional correlations from a single year. They are consistent with the capability story, but they do not establish that complexity causes income: rich countries can afford to build complex industries, so the causal arrow plausibly runs both ways, and both may be driven by institutional quality or historical circumstance. Establishing direction would require panel data and an identification strategy — see Panel Data and Fixed Effects in R and Reasoning with DAGs.

5 Summary

• GDP measures the market value of final output produced domestically. It is well standardised and widely available, and it is not a welfare measure.
• Use real figures for comparisons over time and PPP figures for comparisons across countries; the choice changes the answer.
• GNI measures income accruing to residents. Divergence from GDP flags remittance dependence, profit repatriation, or a large foreign workforce — which is why Ireland’s and Qatar’s GDP figures mislead.
• Economic complexity measures productive capabilities rather than income flows, and correlates with income independently of natural resource wealth.
• All three describe production and income. None describes distribution, which is the subject of Measuring Inequality and Poverty, or human capabilities, covered in Human Development and Social Indicators.

References

Aistleitner, Matthias, Claudius Gräbner, and Anna Hornykewycz. 2021. “Theory and Empirics of Capability Accumulation: Implications for Macroeconomic Modeling.” Research Policy 50 (6): 104258. https://doi.org/10.1016/j.respol.2021.104258.

Hidalgo, César A. 2021. “Economic Complexity Theory and Applications.” Nature Reviews Physics 3 (2): 92–113. https://doi.org/10.1038/s42254-020-00275-1.

Hidalgo, César A., Bailey Klinger, Albert-László Barabási, and Ricardo Hausmann. 2007. “The Product Space Conditions the Development of Nations.” Science 317 (5837): 482–87. https://doi.org/10.1126/science.1144581.

Mankiw, N. Gregory. 2024. Principles of Economics. 10th ed. Cengage.

Footnotes

1. You will sometimes encounter GDI (Gross Domestic Income), which measures the same activity from the income side and should in principle be identical.

2. On capabilities and their accumulation see Aistleitner et al. (2021); on the complexity measure itself, the Atlas FAQ or, at an academic level, Hidalgo (2021).

3. The original article remains worth reading (Hidalgo et al. 2007).