Quantum mechanics

Process (computing)

Operating system

Developmental psychology

Economics

Survivability

Computer science

Dropout (neural networks)

Computer network

Physics

Nonlinear system

Psychology

Mathematics

Econometrics

Productivity

Maturity (psychological)

Economic growth

Machine learning

Scientometrics

Survivability is one of the features for success in contemporary science ecosystem. In this paper, we analyze the publication records of physicists in American Physical Society journals, aiming to identify the career length of each researcher and accordingly investigate the dropout phenomenon in science by the example of physicists. We find that scientific career is a complex nonlinear evolution process and can be generally divided into four stages regarding the dropout rate. In the early career, the dropout rate from trainee phase to maturity is high and negatively correlated with the research performance of the scientists, in both productivity and impact. Moreover, a strong nonlinearity is observed when we study the detailed relationship between the dropout rate and research performance. Interestingly, in the more mature stage of the career, the dropout rate becomes stable and independent of the early performance of the scientists. In the late career stage, the dropout rate increases and is mainly determined by retirement and external factors. The findings in this paper may provide useful guidance for young scholars to allocate their research effort in the early career.

The strong nonlinear effect in academic dropout