A telling graph

This graph shows how much the current Australian government like universities. They are giving subsidy to businesses so much so that the profit is up compared to pre-Covid, but they don’t find a cent to help universities.

 First-half 2020 Business Profits and

In memory of my dearest mother

1938 – Nov 2020

Medical “facts” you probably didn’t know

Probably about 10% of COVID-19 cases lead to 80% of the spread.

About 8 percent of human genome is made of sequences that originated as invasive viruses. To put that number in perspective, genes make up about 1 to 1.5 percent of your genome.

The dismal state of progress in cancer treatment (according to Azra Raza, treatment for 32 percent of cancers patients is essentially the same as 50 years ago)

Some Interesting Stats

Consumer Spending is now down only by 3 percent

Credit application volumes have fallen around 30%

According to The Australian Banking Association, 429,000 mortgages have been deferred totalling $153.5 billion. The total number of loans deferred to 703,000, worth $211 billion.

ICP-2017 report is out

World Bank’s 2017 ICP (International Comparison Program) results are out. Some of the findings reported below are taken from here. The full report can be accessed from here.

  • In 2017, the size of the world economy was nearly $120 trillion as measured by PPPs, compared to nearly $80 trillion as measured by market exchange rates. The two largest economies in the world in 2017 were China and the United States, each of whom recorded a PPP-based GDP of just under $20 trillion. Together, they accounted for a third of the global economy.
  • High-income economies accounted for nearly half of PPP-based global GDP in 2017 while upper-middle-income economies produced just over a third. Lower-middle- and low-income economies accounted for 16% and 1% of the global total respectively. 
  • Around three quarters of the global population lived in economies where the average GDP per capita in 2017 was less than the global mean for that year of $16.6K. Luxembourg had the highest GDP per capita at $112.7K.  Qatar and Singapore followed, both recording a GDP per capita of more than $90K. 
  • At the same time three quarters of people lived in economies where consumption per capita was below the $10.9K global average in 2017. The United States had the highest consumption level at $44.6K per capita. 
  • When comparing price levels, Bermuda was the most expensive economy in 2017, followed by Iceland, Norway, and Switzerland. The cheapest economy was Egypt, followed by Ukraine, Sudan, and Kyrgyzstan. 
  • The United States had the highest health expenditures per capita at $9.4K in 2017, while Germany had the second highest at $6.2K. 
  • At 26%, China contributed the largest share of the world’s expenditures on investment in 2017, followed by the United States at 14%.

Hitler’s paper gets rejected

Downfall is a good German war drama movie, some scenes from this movie have been used to create hundreds of funny Youtube videos on various topics. The following video is quite funny and under-viewed (warning: the video has swearing in it)

Some Links

Universities and Pandemic (John Quiggin)

COVID-19’s Impact on Australian Research Workforce 

Australian economists back social distancing measures

Intergenerational Mobility in US (Chetty vs Heckman)

Modelling of COVID-19 Dynamics

In the first lecture of my economics classes, I often lecture students on how good studying economics is. I tell them that economists study big and cool questions, and not only that, they often give respectable answers (sometimes even to questions out of their field) and that is because they are well-trained in both modelling and data analysis. I am careful enough not to claim that they give a correct or true answer, true answer is a very high bar. I don’t know if students believe me or they think that I am just self promoting but works such as this one might be a case in point (at least in the hand of top-rated economists).

In this study, Jesús Fernández-Villaverde and Charles I. Jones model the dynamics of COVID-19 and they appears to improve over the main model used by epidemiologists. The so called SIR (Susceptible, Infectious, or Recovered) model is the basic model used by epidemiologists to study the dynamics of infectious diseases. One of the important parameters in this model is R0 (basic reproduction ratio) which is defined as the expected number of new infections resulting from one infected person. In the basic model, R0 is assumed to be fixed over time which could lead to bad forecasting due to a “Lucas Critic” sort of argument. As I understand, the novelty in their model is to allow R0 to be time-varying.

Two interesting results from their paper that I noticed are: (1) It is not possible to estimate an accurate measure of death rate for the disease from such models. Different death rates (e.g. 0.3%  and 1%) produce essentially the same observed dynamics although their long run impacts and policy implications are very different. (2) Accurate forecasting of the number of death in the early stages of a novel virus (before its peak) may not be possible but after the peak forecasts seem to perform well as this figures from their paper shows. For New York which has passed the peak, the 7 forecasts (by adding daily data for seven consecutive days) are all converging but for California they produce very different forecasts.


Also check out their estimates of cumulative number of death for US and many other countries from here

Off the Charts

The following chart is taken from the New York Times article:  How Bad Is Unemployment? ‘Literally Off the Charts’


MDD using VB

In this paper (joint with my colleague from Melbourne Uni) we show that the variational Bayes density is an ideal candidate for calculating marginal likelihood. The most recent version can be obtained from here

Accurate Computation of Marginal Data Densities Using Variational Bayes

Gholamreza Hajargasht and Tomasz Wo’zniak

Abstract: We propose a new marginal data density estimator (MDDE) that uses the variational Bayes posterior density as a weighting density of the reciprocal importance sampling (RIS) MDDE. This computationally convenient estimator is based on variational Bayes posterior densities that are available for many models and requires simulated draws only from the posterior distribution. It provides ac-curate estimates with a moderate number of posterior draws, has a finite variance and provides a minimum variance candidate for the class of RIS MDDEs. Its reciprocal is consistent, asymptotically normally distributed, and unbiased. These properties are obtained without truncating the weighting density, which is typical for other such estimators. Our proposed estimators outperform many existing MDDEs in terms of bias and numerical standard errors. In particular, our RISMDDE performs uniformly better than other estimators from this class