As many people know too well, it has not been a good year for jobs in pharma and biotech. As STAT News reported, it is hard out there for job seekers.
I have previously done some data mining for pharma job postings, so I went ahead and browsed more recent data to understand whether data science jobs might have been affected more or less than other roles.
My methods are imperfect, but I focused on top pharma and jobs descriptions that focused on data science and related-skills.
Overall, jobs are down
I used data from November 2022, February 2023, and November 2023 (when I downloaded the job descriptions).
Show code
library(dplyr) # data wrangling
library(gt) # make nice tables
library(ggplot2) # nice plots
# plot theme
theme_set(theme_minimal(base_family = "Avenir"))
# set seed
set.seed(1203)
data_files <- c("2022-11-03/combined_data2.csv",
"2023-02-20/combined_data2.csv",
"2023-11-21/combined_data2.csv")
jobs <- purrr::map_dfr(
file.path("~/Documents/code/job_description_collector/data/", data_files),
read.csv, .id = "id")
date_map <- data.frame(
id = as.character(1:length(data_files)),
date = dirname(data_files)
)
jobs <- jobs |>
left_join(date_map, by = "id")
# require distinct descriptions
jobs <- distinct(jobs,
date, company, title, description, .keep_all = TRUE) |>
add_count(date, company, name = "n_company")Looking at all jobs, we can see that the postings have decreased from February to November.
Show code
jobs |>
distinct(date, n_company, company) |>
ggplot(aes(date, n_company, group = company)) +
geom_line() +
geom_point(aes(color = company)) +
coord_cartesian(ylim = c(0, NA)) +
labs(title = "Total jobs per company",
y = "Jobs", x = "Date")
There are few data science jobs
Next, I labeled jobs as “data jobs” if the descriptions had R or data science in them. Most the jobs are not “data jobs,” so to look at the trends, I normalize the values to the first data point (November, 2022).
Back in November of 2022, there were 232 jobs that mentioned R or “data science”. In November of 2023, that is down to only 66 jobs.
I also plotted “chemistry” and “sales” as comparisons. It seems that some types of roles have been hit harder than others!
Show code
jobs_w_type <- mutate(jobs,
job_type = case_when(
grepl("[^(Maurice)] R[\\., ][^(&N.Ph)]", description) |
grepl("[Dd]ata [Ss]science", description) ~ "data",
grepl("[Cc]hemistry", description) ~ "chemistry",
grepl("[Ss]ales", description) ~ "sales",
TRUE ~ "other")
) |>
count(job_type, date) |>
group_by(job_type) |>
mutate(fraction = n / n[date == "2022-11-03"])
jobs_w_type |>
ggplot(aes(date, fraction, group = job_type, color = job_type)) +
geom_line() +
geom_point() +
coord_cartesian(ylim = c(0, NA)) +
scale_y_continuous(labels = scales::percent) +
scale_color_manual(values = c("#87B8DA", "firebrick4", "#87B8DA", "#87B8DA")) +
geom_text(aes(label = job_type), nudge_x = 0.1, hjust = 0,
data = filter(jobs_w_type, date == "2023-11-21"),
family = "Avenir") +
labs(title = "Jobs by type",
y = NULL, x = "Date",
color = "Job type") +
theme(legend.position = "none")
Good luck out there job seekers! If anyone is looking for job search advice, feel free to reach out to see if I can help.
sessionInfo
pander::pander(sessionInfo())R version 4.3.0 (2023-04-21)
Platform: aarch64-apple-darwin20 (64-bit)
locale: en_US.UTF-8||en_US.UTF-8||en_US.UTF-8||C||en_US.UTF-8||en_US.UTF-8
attached base packages: stats, graphics, grDevices, utils, datasets, methods and base
other attached packages: ggplot2(v.3.4.4), gt(v.0.10.0) and dplyr(v.1.1.2)
loaded via a namespace (and not attached): gtable(v.0.3.3), jsonlite(v.1.8.4), highr(v.0.10), compiler(v.4.3.0), Rcpp(v.1.0.10), tidyselect(v.1.2.0), xml2(v.1.3.4), jquerylib(v.0.1.4), scales(v.1.2.1), yaml(v.2.3.7), fastmap(v.1.1.1), R6(v.2.5.1), labeling(v.0.4.2), generics(v.0.1.3), knitr(v.1.42), tibble(v.3.2.1), pander(v.0.6.5), distill(v.1.6), munsell(v.0.5.0), bslib(v.0.4.2), pillar(v.1.9.0), rlang(v.1.1.1), utf8(v.1.2.3), cachem(v.1.0.7), xfun(v.0.39), sass(v.0.4.5), memoise(v.2.0.1), cli(v.3.6.1), withr(v.2.5.0), magrittr(v.2.0.3), digest(v.0.6.31), grid(v.4.3.0), rstudioapi(v.0.15.0), lifecycle(v.1.0.3), vctrs(v.0.6.4), downlit(v.0.4.2), evaluate(v.0.20), glue(v.1.6.2), farver(v.2.1.1), fansi(v.1.0.4), colorspace(v.2.1-0), purrr(v.1.0.2), rmarkdown(v.2.21), tools(v.4.3.0), pkgconfig(v.2.0.3) and htmltools(v.0.5.5)