Institutions and policies of economic freedom: different effects on income and growth

Abstract

The aim of this paper is to provide some additional results concerning how economic freedom enhances growth and/or long-run income. Our hypothesis is that institutions and policies of economic freedom may have different effects on long-run income and growth, both in terms of their size and working mechanisms. To test our hypothesis, we apply both Acemoglu et al.’s (Am Econ Rev 91(5):1369–401, 2001) and Mankiw et al.’s (Q J Econ 107(2):407–437, 1992) modeling strategies in our cross-country regression analyses. The major finding is that the institutions of economic freedom are of primary importance in economic development: they matter both in the long run and during the catching up period, and have both direct and indirect effects. On the other hand, the effects of economic freedom policies (monetary and fiscal) matter only during the catching up period, with the fiscal policy having a more straightforward effect. Furthermore, the estimates from the two modeling strategies are in line with each other, which can be seen as a further corroboration of our results.

Appendix

1.1 A comparison of the long-run and the short-run estimates

In Sect. 4.4 we made four assumptions in order to be able to compare our short-run estimate with the long-run one. The first two assumptions can be summarized by saying that the change in log income can be written as

$$\Delta \ln y_{2010} = \sum\limits_{t = 1801}^{2010} {\hat{\alpha }_{5} \Delta \ln (inst)_{t} }$$

where Δln(inst) _t is the change in the economic freedom institutions in year t, and $\hat{\alpha }_{5}$ is the estimation for the effect of economic freedom institutions on the growth rate we derived from the analysis of the period 1980–2010.

The third assumption we made concerns the time series economic freedom institutions follow in the long-run. To be able to make a meaningful estimation of the long-run effect from the short-run one, we have to consider what happens to the measure of economic freedom institutions in the periods we are not able to observe. The regression results on the growth rate can be understood as what would happen if the measure of economic freedom institutions (ln institutions) was increased by one unit at the beginning of a 30-year period and remained at this higher level until this period ends. This 30-year period is just the last 30 years of the long period in question (210 years), and a one-unit difference for this last 30 years does not imply the same difference for the preceding 180 years. It is therefore meaningful to divide this long period into two parts:

$$\Delta \ln y_{2010} = \sum\limits_{t = 1801}^{1980} {\hat{\alpha }_{5} \Delta \ln (inst)_{t} } + \sum\limits_{t = 1981}^{2010} {\hat{\alpha }_{5} \Delta \ln (inst)_{t} } = \hat{\alpha }_{5} \left[ {\sum\limits_{t = 1801}^{1980} {\Delta \ln (inst)_{t} } + 30} \right]$$

To be able to make an educated guess about the difference in economic freedom institutions before 1980, we assume a time series which institutions follow. Suppose that this process is a stationary one for country j as follows:

$$\ln (inst)_{jt} = \theta \ln (inst)_{jt - 1} + \varepsilon_{jt}$$

with θ being the autoregressive parameter to be estimated from the long-run data of economic freedom, while ε _t is the error term. Therefore, the difference in economic freedom institutions between two countries follows the process:

$$\Delta \ln (inst)_{t} = \theta \Delta \ln (inst)_{t - 1} + \varepsilon_{t}$$

Our assumptions above say that this difference is zero in 1800 and it is one in 1980. To make use of this, first note that the difference in a year t (1801 ≤ t ≤ 2010) is

$$\Delta \ln (inst)_{t} = \sum\limits_{k = 1801}^{t} {\theta^{t - k} \varepsilon_{k} }$$

Second, let us suppose that the one-unit difference observed in 1980 is the result of a slow and steady evolution of the institutions of economic freedom through which the same shock occurs each year:

$$\varepsilon_{1801} = \varepsilon_{1802} = \cdots \cdots = \varepsilon_{2010} \equiv \varepsilon$$

Thus, assuming a one-unit difference for 1980 implies assuming a value for ε because $\Delta \ln (inst)_{1980} = \sum\limits_{k = 1801}^{1980} {\theta^{1980 - k} \varepsilon = } \sum\limits_{l = 0}^{179} {\theta^{l} \varepsilon = } \frac{{1 - \theta^{180} }}{1 - \theta }\varepsilon = 1$, that is $\varepsilon = \frac{1 - \theta }{{1 - \theta^{180} }}$.

In this way the expression for the income differences becomes:

$$\Delta \ln y_{2010} = \hat{\alpha }_{5} \left[ {\sum\limits_{t = 1801}^{1980} {\Delta \ln (inst)_{t} } + 30} \right] = \hat{\alpha }_{5} \left[ {\sum\limits_{t = 1801}^{1980} {\sum\limits_{k = 1801}^{t} {\theta^{t - k} \varepsilon } } + 30} \right]$$

$$= \hat{\alpha }_{5} \left[ {\sum\limits_{t = 1801}^{1980} {\frac{{ \, 1 - \theta^{t - 1800} }}{1 - \theta }} \frac{1 - \theta }{{1 - \theta^{180} }} + 30} \right] = \hat{\alpha }_{5} \left[ {\sum\limits_{l = 1}^{180} {\frac{{1 - \theta^{l} }}{1 - \theta } + 30} } \right] = \hat{\alpha }_{5} \left[ {\frac{180}{{1 - \theta^{180} }} - \frac{\theta }{1 - \theta } + 30} \right]$$

When examining the long-run data of Prados de la Escosura (2016) we find that in quite a few countries among the 21, the relevant component is stationary.¹⁸ Using these long-run data, the value of θ, as averaged over the “stationary countries”, is estimated to be about 0.997, which gives the final result:

$$\Delta \ln y_{2010} = \hat{\alpha }_{5} [98.573 + 30] = \hat{\alpha }_{5} \times 128.573$$

As the estimated value of α ₅ runs between 0.028 and 0.079 in Tables 3 and 4, this gives an estimate between 3.600 and 10.157 for the income difference in 2010, which shows a substantial overlap with—although it is far from being the same as—the interval between 2.476 and 4.600, the range within which we estimated the effect of economic freedom institutions on income in Table 2.

The time series of economic freedom institutions might not be stationary, however. Using panel unit root tests and considering the overall index, rather than any of its individual components, Sobel and Coyne (2011) came to the conclusion that economic freedom follows a non-stationary process. As we saw,¹⁹ even the components of economic freedom institutions found in the Prados de la Escosura (2016) database follow a non-stationary process in some of the 21 countries. In accordance with this, let us derive a similar estimate to the one made above, assuming a random walk process for economic freedom institutions:

$$\ln (inst)_{jt} = \ln (inst)_{jt - 1} + \varepsilon_{jt}$$

leading to the difference between our two hypothetical countries in the form:

$$\Delta \ln (inst)_{t} = \sum\limits_{k = 1801}^{t} \varepsilon = (t - 1800)$$

For 1980, this becomes

Δln(inst)₁₉₈₀ = 180ɛ = 1, implying that ɛ = 1/180.

With the help of this we can again derive an educated guess for the income difference:

$$\Delta \ln y_{2010} = \hat{\alpha }_{5} \left[ {\sum\limits_{t = 1801}^{1980} {(t - 1800)\varepsilon } + 30} \right] = \hat{\alpha }_{5} \left[ {\sum\limits_{l = 1}^{180} {l\varepsilon } + 30} \right] = \hat{\alpha }_{5} \left[ {180 \times \frac{180 + 1}{2} \times \frac{1}{180} + 30} \right]$$

Which, finally, is

$$\Delta \ln y_{2010} = \hat{\alpha }_{5} [90.5 + 30] = \hat{\alpha }_{5} \times 120.5$$

Using the numbers from Tables 3 and 4 we now end up with an interval of 3.374–9.520 as an alternative estimate of what we have already estimated in Table 2. This interval shows a slightly greater overlap with our estimation.