RDP 2024-01: Do Monetary Policy and Economic Conditions Impact Innovation? Evidence from Australian Administrative Data 6. Macroeconomic Effects on Productivity

Thus far we have focused on measures of innovation rather than productivity. Moran and Queralto (2018) demonstrate in their model and empirically that shocks to R&D spending have an effect on productivity in the medium run. They show this focusing on the United States and using a cross-country panel (that includes Australia). However, given the vastly different structures of the Australian and US economies, it is worth examining whether the results for the United States hold when focusing only on Australia.

To examine if their results hold for Australia specifically, we reproduce the small VAR model used in Moran and Queralto (2018). We estimate a small, three variable VAR with annual data on GDP, TFP from Bergaud, Cette and Lecat (2016), and R&D spending from 1988 to 2019. As in Moran and Queralto (2018) we examine the effect of R&D shocks, which are identified by ordering R&D last in a Cholesky decomposition. This implies that R&D cannot affect TFP contemporaneously, consistent with the fact that it generally takes time for R&D expenditure to result in new technologies. While the assumption used to identify may still be strong, given we are using the same identification as Moran and Queralto (2018), it will give a direct Australian comparison to their US results.

Taking this approach, we find that a positive R&D shock leads to a persistent increase in TFP that peaks around five years after the shock (Figure 2). The magnitudes of the responses are somewhat larger than Moran and Queralto (2018) report, with an equivalent size increase in R&D (4 per cent) leading to a 1.6 per cent increase in TFP, compared to 0.4 per cent in Moran and Queralto. In the data, the volatility of the R&D shock is an order of magnitude smaller (around ½ per cent, compared to 4 per cent in Moran and Queralto (2018)), suggesting that such a large shock would be extremely unusual.^[13] Again, our results are somewhat more in line with Ma and Zimmerman (2023). Drawing on other estimates, they suggest that a 100 basis point shock would lower TFP and output by 0.5 to 1 per cent, whereas our estimates put this around 1.6 per cent.

Interestingly, the response is also less long-lived compared to Moran and Queralto (2018): the peak occurs around five years after the shock for Australia, compared with around eight years in the United States. This could potentially reflect the fact that Australia imports innovation, and so while adoption of technologies declines the actual stock of global knowledge is unaffected, allowing firms to catch up more quickly (though still with a substantial lag).

Overall, while the exact magnitudes differ somewhat to Moran and Queralto (2018), these results suggest that changes in R&D and innovation can have long-run effects on productivity in Australia, as is evident in the United States.

Footnote

We experimented with looking directly at the effects of monetary policy shocks on productivity. We were unable to identify any effects. However, this likely reflects the lack of quarterly data on TFP, resulting in our estimates being based on a small number of observations with annually aggregated shocks. [13]