Prediction algorithms can improve decision-making

In his doctoral dissertation at Hanken School of Economics, Joosua Virtanen develops and applies prediction algorithms in two areas: firm growth and public procurement.

“Economic decisions are full of prediction problems,” Virtanen explains. “Yet economists often focus on the causal effects of explanatory variables to outcomes rather than asking how well we can predict outcomes. I wanted to show what happens when we take prediction seriously: why it matters and how much it matters”

Predicting which companies will grow rapidly has long been considered nearly impossible. Virtanen’s research challenges that assumption. Using detailed data on Finnish companies, he demonstrates that machine learning models altered with an economic principle of a budget constraint significantly outperform traditional prediction methods.

“When venture capitalists have limited resources, they need to choose wisely,” Virtanen says. “My research together with co-authors shows that applying simple economic concepts to data driven methods can help in this task tremendously”

Virtanen also studies public procurement, a sector that represents a large share of GDP but is infamous for inefficiencies, and prone to favoritism. Virtanen and co-authors introduce a new way to detect favoritism by analysing bidding patterns.

“We found that you can identify favoritism by looking at discontinuities in past wins between close winners and losers,” he explains. “This method, borrowed from econometrics, is repurposed here to uncover patterns that suggest potentially illegal practices.”

Tests using Finnish cleaning service auctions revealed bidding behaviors consistent with preferential treatment. The method could help improve transparency and efficiency in public spending.

In total, the dissertation consists of four essays, two of which consider prediction problems in firm growth and two in public procurement.

You can read the whole thesis here:
Essays on microeconomic prediction problems: with applications to firm growth and public procurement

Joosua Virtanen will defend his doctoral thesis on 26 November at 12:00 at Hanken School of Economics, Arkadiankatu 22, Helsinki.

Opponent: Johan Lundberg, Umeå University
Custos: Ari Hyytinen, Hanken School of Economics

You can participate in the defence on-site or via this Teams link: 
Defence Virtanen

More information:

Joosua Virtanen
Phone: 040 8460 677
E-mail: joosua.virtanen@hanken.fi