Lukas Püttmann    About    Research    Blog

Old and not-so-rusty: Successful founders are middle-aged


When thinking of successful startup founders, many people have Mark Zuckerberg, Steve Jobs or Bill Gates in mind - in general a young college-dropout working from a garage in Silicon Valley.

In a new paper (pdf) Pierre Azoulay, Benjamin F. Jones, J. Daniel Kim and Javier Miranda attack this misconception of young founders being more successful than older ones.

The authors take comprehensive census data for the United States and show that the average age for successful founders is in the 40-45 year range. As they write:

The mean age at founding for the 1-in-1,000 fastest growing new ventures is 45.0.


They show in clever ways that this unexpectedly high age average also holds for high-growth startups and those in the tech industry, by using ex ante and ex post measures of growth and tech specialization.

Ex ante, they measure the industry a founder is in, whether the firm received venture capital (VC) funding and whether the firm holds a patent. Ex post, they are able to measure employment and sales growth for these firms 3, 5 and 7 years after founding and they measure which firms “exit successfully” (meaning they are sold or go public). The average ages are similar for all these different subgroups.

Azoulay et al. also show that the results are not just driven by the fact that middle-aged founders are more common thus pushing up the age average. Instead they show that middle-aged founders are more likely to succeed.

One reason for this might be the importance of prior experience. They show that the probability of success (being one of the 1-in-1,000 fastest growing firms) doubles after three years of work in the relevant industry.

The authors also discuss extreme outliers (Zuckerberg etc.) and provide the following verdict:

With [keeping the anecdotal type of evidence] in mind, however, the patterns may suggest a potential reconciliation between the existence of great young entrepreneurs and the advantages of middle age. Namely, extremely talented people may also be extremely talented when young. These individuals may succeed at very young ages, even when people (including these young successes) get better with age. Thus there is no fundamental tension between the existence of great young entrepreneurs and a general tendency for founders to reach their peak entrepreneurial potential later in life.


Time period: My biggest criticism is that startups only make it into the sample used for the most important analyses if they were founded in 2007-2009. This is a short sample to start with, but more importantly this was also the time of the Financial Crisis.

During this large recession, investors might have stopped funding and banks might have stopped giving credit first to the more speculative and riskier firms run by younger people. This might bias the results to older founders. Table A4 shows averages separately for the different years, but all these years are the aftermath of the Financial Crisis, so that doesn’t address this problem.

You could also think that there might be some cohort effect of all the 40-year olds in those years having been young in the 90s when they might have started their first companies during the first Dotcom gold rush.

Measures of success: A second criticism I have is that this paper defines employee and revenue growth as the key measures of a startup’s success. The most clean measure would is a startup’s value, which is a more complete summary of a firm’s ability to generate profits in the future. This is difficult to measure even for established firms, so I understand why the authors did not do this.

However, this can lead to extreme examples where the author’s measures make less sense. Take the example of WhatsApp: It had only 50 employees in early 2013, but was acquired by Facebook a year later for 19 billion USD. (To be fair, with 400 million users in late 2013 and each paying a dollar for the product a year, revenues would already have been impressive).

On balance: But in total, I think this is an excellent paper. It hits a sweet spot of 1) discussing a relevant topic, 2) bringing novel results that shift my views and 3) being well-executed and thus trust-worthy.

I also find it somewhat comforting to think that you don’t have to rush into entrepreneurship with a half-baked idea, but that there are advantages to starting a firm later in life.

Playing with the data

Most of the data is confidential, so the authors can’t share it. But they’ve also collected a list of companies named by media and VC organizations that were considered particularly promising and the age of their founders. This data we can play with.

Open up R and import these libraries:


Instead of manually downloading the files from the AER website and saving them somewhere, we can download them temporarily and just keep what we need:

temp <- tempfile()
download.file("", temp, mode="wb")

The data we need are stored in four files: Matrix.dta, Sequoia.dta, TechCrunch.dta and Magazines.dta. Instead of downloading each and combining them, we can do this only writing the code once.

This defines a function which unzips the relevant file, imports the Stata files and adds a column to describe the data source organization:

import_org_dta <- function(org_name) {
  unz(temp, paste0("programs/", org_name,".dta")) %>% 
    read_dta() %>% 
    mutate(org = org_name) %>%
    select(org, everything())

Use this function on the list of input names:

orgs <- c("Matrix", "Sequoia", "TechCrunch", "Magazines")

perceptions <- lapply(X = orgs, FUN = import_org_dta) %>% 

Close tempororary connection:


The resulting dataset perceptions holds the name of the organization, the name of the list they created, the age of the founder and the company’s name:

> perceptions

# A tibble: 944 x 4
#   org    list             age company_name  
#   <chr>  <chr>          <dbl> <chr>         
# 1 Matrix Matrix Partne…    45 Akorri Networ…
# 2 Matrix Matrix Partne…    36 Atria Software
# 3 Matrix Matrix Partne…    33 Atria Software
# 4 Matrix Matrix Partne…    33 BlueMartini   
# 5 Matrix Matrix Partne…    32 Canva         
# 6 Matrix Matrix Partne…    25 Canva         
# 7 Matrix Matrix Partne…    27 Canva         
# 8 Matrix Matrix Partne…    48 Clarify       
# 9 Matrix Matrix Partne…    62 Clarify       
# 10 Matrix Matrix Partne…    43 Clarify       
# ... with 934 more rows

Plot the age of the founders and add a line marking the paper’s estimate of 45 years against it:

ggplot(perceptions, aes(age, org, color = list)) +
  geom_jitter(alpha = 0.8, stroke = 0, size = 2.2) +
  labs(title = 'Founder Age – Perceptions from Media & Two Prominent VCs',
       subtitle = "Vertical line marks estimate from study (age = 45)",
       caption = 'Source: "Age and High-Growth Entrepreneurship" (2020) 
       by Azoulay, Jones, Kim and Miranda.', y = NULL, x = "Founder age", color = "List: ") +
  theme_minimal() +
  geom_vline(xintercept = 45) 
Typical startup founder age as portrayed by two media sources and two prominent venture capital firms

If the public perception of founders’ age was in line with the estimates from this paper, we would expect the cloud of point to be centered around the black line.

The fact that these points lie mostly to the left of the line shows that public perception favors younger founders.


Azoulay, Pierre, Benjamin F. Jones, J. Daniel Kim and Javier Miranda (2020). “Age and High-Growth Entrepreneurship.” American Economic Review: Insights. (doi)

Further reading: Related posts: