The Cournede-Denk Results

Cournede and Denk use data for 33 OECD countries for the period 1961­2011 to estimate cross-country growth regressions incorporating a linear term on finance. I will focus on their results for “intermediated credit,” credit to the private sector from either banks or other financial institu- tions.^[217] Table 6.1 reports their coefficient estimates for three specifications of regressions for per capita GDP growth.

The size of the negative impact of credit is extremely large. In the second and third columns, this coefficient is -0.019. This magnitude means that if credit to the private sector were reduced from 150 to 100 percent of GDP, the annual growth rate per capita would increase by 0.95 percent. If credit for such a country were eliminated altogether, then according to this linear equation the annual growth rate would rise by nearly 3 percentage points. Both the linearity (and hence optimal level of zero for credit) and the mag­nitude of the impact are implausible.

Nor are these results consistent with saying that, although by now the OECD has gone too far in finance, at earlier periods of lower financial depth more finance meant more growth, even though the authors seek to argue that this is the case. The absence of a quadratic term and the presence of a negative coefficient on the linear term for finance mean that additional finance reduces growth across the entire period and all OECD countries.

Thus, on average, Australia, Belgium, Greece, Israel, and New Zealand had private credit of only 14.5 percent of GDP in 1961-65. By 2001-05 the average for these countries had reached 83 percent of GDP. Applying the coefficient of -0.019 in the second column of 6.1, the main results reported by Cournede and Denk would imply that rising financial depth in these five countries damaged their growth performance by an average of 1.3 percent per year.^[220] By this diagnosis, the five countries' GDP levels would have been 112 percent higher than their actual levels in 2001-05 if they had kept finan­cial depth frozen at the low levels of 1961-65.^[221] Such a diagnosis, however, is not credible. The basic problem is that the authors reported as their main findings results that differ from their true judgment. Instead, they should have conducted and reported tests that distinguish between economies at low versus high levels of financial depth.^[222]

Even if one were to distinguish between countries starting at higher levels of finance and those starting at lower levels, the estimated magnitudes of growth impact are implausible. Consider Japan and the United States, where private credit rose from a range of 70 to 80 percent of GDP in 1961-65 to 180 to 200 percent in 2001-05. Already in 1961-65 they would have been unnecessarily sacrificing about 1.4 percent in annual growth because of excess finance, according to the equation in the third column of table 6.1; by 2001-05, they would have been sacrificing 3.6 percent annual growth.^[223] But if one adjusts this estimate because the negative effect really only would begin at finance of 60 percent of GDP (a threshold identified by the authors as discussed below), the resulting shift would still mean that by 2001-05 the two countries would have been sacrificing about 2.5 percent annual growth as a consequence of too much finance.^[224] Such a loss would have meant that in the absence of financial deepening, their labor productivity growth by 2001-05 could have been an implausibly high 4.1 to 4.5 percent per year instead of the actual pace of 1.6 to 2.0 percent (OECD 2015a).

Another central problem with the estimates is that they do not apply the most important variable typically included in cross-country growth analysis: the logarithm of real ppp per capita income. Some tests include per capita income but it is instead for the specific country and not comparable across countries.^[225] Testing cross-country growth patterns without permit­ting a comparable cross-country level of real per capita income is a classic instance of staging Hamlet without the Prince of Denmark.

The authors explicitly include “country fixed effects,” equivalent to a dummy variable for each country, and thereby throw out potentially im­portant cross-country information. In the tests conducted here, it turns out that this important decision is responsible for turning the influence of finance from positive to negative. Yet it is arguably inappropriate, and at best the case for including country fixed effects is ambiguous.