3.3 - Uneven globalisation and transnational firms

Measuring globalisation

Globalisation has several dimensions, so the choice of indicator changes the pattern that is visible.

A country's trade as a percentage of GDP measures one economic connection. International visitor numbers, migrant links, information traffic and international agreements reveal different social or political connections. A large country may have enormous absolute trade but a smaller trade-to-GDP ratio than a small open economy. GDP is the value of output produced within an economy; a ratio compares trade with that scale rather than simply comparing total dollars.

The specification names both the AT Kearney index and the KOF index. The historical AT Kearney/Foreign Policy Globalization Index grouped indicators into economic integration, personal contact, technological connectivity and political engagement, as its 2007 index description explains. Use its results with their year and method; it is not a current ranking interchangeable with KOF.

The KOF Globalisation Index combines economic, social and political dimensions. It distinguishes de facto connections, such as actual trade and migration, from de jure conditions, such as trade rules and investment restrictions. A state may permit connections legally but have little actual flow because infrastructure or demand is weak. Conversely, a small number of strong trading relationships may coexist with restrictive rules. KOF uses a more detailed normalisation and weighting method than the simple classroom index in the next part.

Index values are not percentages of an economy that is “globalised”. A ranking is relative to other countries included and does not show the size of the gaps between them. An overall score can mask a country that is strongly connected economically but less connected politically, or a highly connected capital city alongside peripheral rural areas. Country averages also say little about who controls flows or whether people benefit.

Compare the same indicator definitions and observation years. A publication date can be later than its data year. Missing or estimated data, chosen weights and national boundaries affect results. An index is useful for describing and comparing connections; it cannot by itself establish that globalisation caused growth, wellbeing or inequality.

Ranking and scaling an index

Ranking orders observations, while scaling places indicators with different units on a common numerical range.

The following original illustrative data concern three invented countries in the same year. They are neither KOF data nor a validated globalisation measure. All indicators are oriented so that larger values are treated as greater connection.

For internet use, rank A first, B second and C third. For trade relative to GDP, rank B first, A second and C third. Rank tells us order, but loses the size of the difference: first and second could be separated by one percentage point or forty. A rank-sum would add ranks with a lower total indicating greater connection; tied values need a consistent convention, such as equal mean ranks.

To preserve more information, use simple min–max scaling within this three-country sample:

For A, the internet score is $\frac{80-20}{80-20}\times100=100$. Its trade score is $\frac{60-20}{100-20}\times100=50$. An equally weighted mean gives A $(100+50)/2=75$. B has an internet score of $33.3$ and a trade score of $100$, giving $66.7$ when the mean is rounded to one decimal place using unrounded working. C scores 0 on both. The resulting order is A, B, C.

Weighting is a judgement. If trade receives 80% of the weight and internet use 20%, A scores $0.8(50)+0.2(100)=60$, while B scores about $0.8(100)+0.2(33.333\ldots)=86.7$. B now ranks above A. The evidence has not changed; the priority built into the index has changed.

An indicator whose high value represents weaker connection must be reversed before combining it. If maximum and minimum are equal, this formula cannot distinguish observations. Adding a new extreme country changes the scale and possibly scores, so do not compare separately normalised yearly samples as if the reference range stayed fixed. Record the sample, units, direction, weights and observation year; then test whether a plausible different weighting changes the conclusion.

TNCs and production networks

Transnational corporations both use global connections and create new ones through their investment, purchasing and marketing decisions.