World Happiness Report 2023 90 Specifically, we undertake a ‘cluster analysis’ using a fairly standard statistical algorithm that uses machine learning to find groupings of similar countries based on a set of observable attributes. The algorithm used also “decides” how many groups are needed to best fit the data.30 The core variables that are used to construct these clusters are the same as those that go into the Pillars of Prosperity index that we constructed above.31 As illustrated in Figure 3.2, we find that allowing two distinct dimensions of heterogeneity across countries does a reasonably good job of describing the data. The first dimension (along the x-axis of the figure) broadly captures differences in state capacity and income, while the second dimension (along the y-axis of the figure) captures political violence.32 The clustering algorithm identifies three distinct clusters of countries as illustrated in Figure 3.2, where we have shaded the three groups in distinct colors and identified each country by its standard three-letter country code. It is striking that these three clusters correspond neatly to the groupings suggested by our theoretical approach to state effectiveness, as summarized at the beginning of this section. The weak states in the figure are those shaded in orange and positioned in the negative orthant of Dimension 1 (state capacity/income) and positive orthant of Dimension 2 (civil war). This rhymes well with the idea that they have relatively high levels of civil war and low levels of state capacity and income. Special-interest states are shaded in blue and have intermediate levels of state capacity and income. These countries are situated in the negative orthant of Dimension 2, which represents high levels of repression. China is a particular outlier in this dimension, with exceptionally high repression. Common-interest states are shaded in green and form a particularly tight cluster. Countries in this cluster belong to the positive orthant of Dimension 1 (state capacity/income) and they have values on Dimension 2 (conflict) that hover around zero, which represents low levels of repression as well as civil war. Implications for Well-being – Theory and Evidence In this section, we draw out the implications of the preceding analysis for well-being. Moreover, we show that these implications are consistent with the patterns in the data, when we measure well-being with life satisfaction data from the Gallup World Poll. Finally, we relate these empirical patterns directly to the determinants of well-being highlighted in Chapter 2. Effective states and well-being Our two-dimensional approach to state effectiveness gives ample a priori reasons to believe that peaceful states with larger state capacities are conducive to higher well-being for their residents. Living in an environment with peace conveys direct benefits, even more so when such peace is not dependent on state repression. Below, we connect this to the themes developed in Chapter 2. Strong state capacities may mean higher taxation. But we expect this to be the case only when cohesive institutions and/or values encourage public spending on common interest programs for the provision of healthcare, education, or infrastructure. Similarly, high legal capacity may help to promote freedom, serve as a bulwark against discrimination, enhance economic opportunities for disadvantaged groups, and prevent abuse of market power or raise product and workplace safety. We expect this pattern to manifest itself in cross-country comparisons. That said, looking at cross-country data is more of a suggestive exercise than a method to pin down convincing causal relations. Moreover, if the elements of effective states cluster together, it would be hazardous to give too much prominence to any single element of state capacity or peacefulness. This would amount to treating better performance in that particular dimension as a kind of silver bullet for well-being. Instead, the presence of development clusters emphasizes that many state features go hand in hand in effective states. In drawing conclusions from our analysis of well-being differences across countries, we should also be realistic about time frames. Besley, Dann, and Persson33 stress that clustering patterns are
RkJQdWJsaXNoZXIy NzQwMjQ=