Data Explorers newsletter no. 12

Kia ora koutou

This is an occasional newsletter from Aaron Schiff. The only possible way you got on this list is by signing up on my website. You can instantly unsubscribe, or please forward it to anyone who may be interested.

The economic cost of COVID in 2020

New Zealand’s GDP data for the fourth quarter of 2020 was published yesterday, so now we can try to quantify the economic impact of COVID over a full year. This requires an estimate of what New Zealand’s GDP would have been in a fantasy world where there was no COVID (in New Zealand or anywhere else). Here’s my attempt:

The chart above shows a prediction (and a 95% prediction interval) of New Zealand’s annual real GDP for 2020 in a world without COVID, versus the actual outcome. This suggests that, without COVID, New Zealand’s real GDP would have been between $257 billion and $266 billion (in real 2009/10 dollars). The reality was GDP of $250 billion, so this implies that the economic cost of COVID in 2020 was between $7 billion and $16 billion in 2009/10 dollars, or between around $9 billion and $19 billion in today’s money. 

Just to be clear, this is an estimate of the economic cost of COVID relative to a world with no COVID. It is not an estimate of the cost of New Zealand’s COVID elimination strategy. Based on what has happened in other countries that made different choices, I firmly believe that New Zealand’s economic performance in 2020 would have been even worse and hence the economic costs of COVID would have been even higher if we had not chosen elimination. 

The 2020 prediction is from a simple forecasting model that I trained on the history of New Zealand’s quarterly seasonally-adjusted real GDP from 1990 to 2019. The model is not super-sophisticated (it’s an ARIMA model – R code here) but such models are often alright for short-term predictions of stable trends. 

The chart below shows the quarterly GDP actuals and predictions that make up the annual numbers above. Most of the impacts were obviously in the second quarter when we had the level 4 lockdown. GDP in quarter 3 and quarter 4 was down on the prediction but still within the 95% confidence range of the forecasts for those quarters, so some of the variation from the trend may be due to factors other than COVID.

These numbers are useful for planning for and trying to prevent future pandemics. We now know that a COVID-type pandemic cost us roughly 5% of our economic output over a year even when we responded quite successfully to eliminate it (and it’s not over yet). Given that cost, have we allocated enough resources to future pandemic planning and prevention? 

State highway bar chart

For something less serious, I had a silly idea to turn State Highway 1 into a population bar chart (R code here). Above is what happens when you plot resident populations of SA2 areas as bars extending from the nearest point on State Highway 1 (bars point right or left depending on whether the populated area is east or west of the highway). 

The population data is the 2018 Estimated Resident Population from Stats NZ (under ‘Population’ in NZ.Stat). As usual this was way more fiddly than I expected. For example, some SA2 names were slightly different in the population data and in the geographic data that I needed to join (e.g. Opua (Far North District) and Opua (Far North district) — spot the difference?). Anyway, it was a fun experiment. 

Here’s a few interesting things that have crossed my screens:

A simple guide to using Git/GitHub with RStudio

Patterns, Predictions, and Actions, a free book about machine learning, including good coverage of causality and causal inference. 

RealRisk - translate relative risks that often crop up in medical news stories into more useful absolute risks. 

How to pick more beautiful colors for your data visualizations. 

Exploring rectangle packing algorithms

Thanks for reading! Aaron +64 9 336 1323

Back to home