web analytics

Number of jobs versus employment rate

Because I’m a cranky git who is largely distrustful of economists, I couldn’t let go of a sense of distrust of what was being presented in this graph:


Dan O’Brien on Twitter

The Irish economy has increased employment more in the past four years than in the 70 years from 1926 to 1996.pic.twitter.com/Ps6JFnduH6

I’m suspicious that this is simply “bigging up” Fine Gael and being misleading in doing so.

For me, the number at work on its own doesn’t give a complete picture; particularly over a 90 year period. Ireland’s population has changed significantly over that period of time and can’t be omitted when considering economic performance (I reckon, anyway). On that basis I argue that employment rate is the appropriate indicator of labour market performance for a national economy.

So I’ve taken the various data with which to present the employment rate over the period 1926 to 2016:

Population 15 years and Over (Number) by Sex, Industrial Group,
Employment Status and Year
1926 1936 1946 1951 1961 1966 1971 1981 1986 1991 1996
All Persons All Industries Total at Work 1223014 1235424 1227745 1219722 1052539 1065987 1054839 1137827 1091155 1149080 1307236
  • I’ve then used the population data from CNA15: Population by Age Group, Sex, Year and Statistic to get the population for 15-64 year olds over the same time range and selecting the same years (so that means not selecting 1979, 2002, 2006 or 2011 as the latter three are in the QNHS dataset)
Population (Number) by Sex, Age Group and Year
1926 1936 1946 1951 1961 1966 1971 1981 1986 1991 1996
Both sexes 15 – 19 years 286187 268326 251487 241182 233832 259356 267727 326429 331100 335026 339536
20 – 24 years 240231 254513 231290 202172 158007 185289 215251 276127 286424 266572 293354
25 – 29 years 215401 216785 208188 198421 145377 149317 172993 246053 258439 246321 259045
30 – 34 years 183300 183284 204696 191566 152787 146625 151351 231958 242689 249071 260929
35 – 39 years 175759 192420 191882 200916 166793 154272 149107 193829 229740 237889 255676
40 – 44 years 169085 162905 165597 180326 170293 163362 152729 165924 191751 225683 240441
45 – 49 years 168893 156583 173062 160915 174625 166517 160124 151850 161740 187762 225400
50 – 54 years 162718 152707 137492 162986 157122 164957 159082 149680 147511 156806 186647
55 – 59 years 125713 143441 134859 128848 136123 147064 154847 149606 142215 142549 153807
60 – 64 years 105146 130378 119225 122060 131060 123840 134066 139266 139978 134566 137946
  • Combining the CNKL1 and CNA15 data I get
1926 1936 1946 1951 1961 1966 1971 1981 1986 1991 1996
Number at work 1223014 1235424 1227745 1219722 1052539 1065987 1054839 1137827 1091155 1149080 1307236
Total 15-64 1,832,433 1,861,342 1,817,778 1,789,392 1,626,019 1,660,599 1,717,277 2,030,722 2,131,587 2,182,245 2,352,781
Employment rate 66.7% 66.4% 67.5% 68.2% 64.7% 64.2% 61.4% 56.0% 51.2% 52.7% 55.6%
  • QNHS Table8a-EmploymentRates shows data by quarter so I took a simple average for the year and get
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
60.6 63.6 65.3 65.9 65.3 65.4 66.3 67.8 68.6 69.2 67.4 61.9 59.7 58.9 58.9 60.5 61.7 63.3 63.6

We can then plot this combined data: Primary y-axis show employment rate (a percentage) and the secondary y-axis shows number of jobs and population (all ages) in millions.

(click to enlarge)
Employment Rate and Population 1926 to 2016
The points to note here are:

  • The employment rate and the “numbers at work” chart show somewhat similar trends over the last 20 years but the interpretation should be markedly different.
  • The increase in population which show that comparisons with the early and mid parts of the 20th century shouldn’t ignore the impact of population change i.e. absolute job numbers don’t have context.
  • The employment rate today is back at end of the century levels; we’re barely midway to recovering to peak levels. Since the 2012 minima, the employment rate is only up ~8% over the period to now. That recovery is obviously good news but it’s not earth shattering.
  • That increase in employment rate needs to be viewed against the overall state of the economy as it affects the nation; the health service is teetering towards collapse and we have a homelessness emergency. Vanity metrics fed to the people in this manner are misleading at best, in reality insulting to people who know the recovery is fragile and not omnipresent.

Cube Travel SL Review

I originally put up a quick video review on the day I brought the bike home from the bike shop and it follows. However, now that I’ve put up some mileage (~750km) I thought it was time to get into a bit more detail. The bike comes in at ~€1,379; additional cost due to my customisations are shown below.

If you watched the video you’ll see that I was quite excited by the bike. Since then, nothing’s changed: I love this bike and it’s perfect for me right now.

Continue Reading


I posted this on Twitter but I also wanted to add it here because it’s important and there seem to be an awful lot of product people who don’t understand the concept of MVP:

10 Ways To Test If Your MVP Is Actually Viable – ReadWrite

Throwing together a minimum viable product, or MVP, is a great way to test your concept and find out what customers really want before you sink more money into the game. However, there’s a fine line between tossing something together and a MVP that’s too buggy to launch.


EPG and recording time trouble in TVHeadend

I’ve moved from having a full XBMC Kodi install on my Raspberry Pi to just Raspbian with tvheadend as all I want is to capture content for consumption elsewhere e.g. RTÉ Pulse shows.

Recently, I had the bright idea of moving to Jessie and in the process discovered that all no longer played well and I wasn’t recording the actual shows I wanted. Two things were happening: the EPG which is pulled from EIT OTA was +1 hour out – most discussions on this issue being observed on TVH centre around there being an issue with the provider. So I went down the rabbit hole of trying to get DVB scanner tools to work to inspect the Saorview EIT data – a successful endeavour. It was then by chance that I saw TVH when debug mode was enabled was logging events with a -1 hour difference.

This is where how time is handled on Linux gets confusing. I looked at the timezone per user (so setting via .profile) and at system level which I naively assumed was /etc/timezone. I wasted lots of times between these two until I discovered /etc/localtime and then this information: https://wiki.debian.org/TimeZoneChanges. dpkg-reconfigure tzdata is your friend.


Custom Gitlab to Slack notifications with Zapier

We’re back to using Slack for product development communication again, so I’ve been keen to tie in the documentation updates I do via Dexy & Asciidoctor to Slack. What I wanted was to notify the channels setup per product of a new version of the specification along with the URL to the documentation site.

Gitlab has Slack integration ‘out of the box’ as described here. Here’s what that looks like:

GitLab [10:30 PM]
James Gallagher pushed to branch master of etckeeper/gitlab (Compare changes)
629601425: saving uncommitted changes in /etc prior to apt run – James Gallagher
ea40dff39: committing changes in /etc after apt run

Package changes:
-adduser 3.113ubuntu2 Show more…

This works great where the audience for the notification is interested in code/configuration changes but doesn’t fit with the use case I described above. So that meant looking instead at a specific Gitlab use case: Web Hooks for tag push events. As I mentioned in the Dexy & Asciidoctor post I tag new versions of specifications – hence tying the notification to ‘tag push events’.

The rest of this post describes the high level approach I took and the [awful] code I used.
Continue Reading


Example for Dexy & Asciidoctor

In the post Dexy & Asciidoc(tor): A Business Analyst’s documentation tool to look at I went through how I’m using the two tools at the moment. I promised to provide an example within a few days but this has turned into a month now. Anyway, attached this post are the HTML (image not included, you’ll need the ZIP) and PDF outputs along with the full Dexy ‘project’. The content is in a Gitlab git repository.


Dexy & Asciidoc(tor): A Business Analyst’s documentation tool to look at

Like many a BA, I get cranky with Microsoft Word as a tool for writing requirements specification documents. ‘Track Changes’ functionality isn’t version control and very quickly becomes unwieldy. I also find that document formatting behaves inconsistently when you get into different sets of section formatting, even when you try to bend numbered lists to your will. I’ve also used IBM’s Rational DOORS in the past for documenting requirements and you get used to that idea of structured editing and establishing relationships between requirements. So, I decided one evening that I’d go to good old plain text and use a markup language for formatting so that I could have absolute control over my documentation.

which led to

Off I went with "What is dexy?" and the bit that stood out for me was:

Dexy makes it easier to create technical documents by doing the repetitive parts for you. Dexy provides a consistent interface to tools and scripts so you don’t have to run them manually. Your project’s dexy configuration keeps track of what to run, in which order, and with what parameters. This way, your whole process is captured so anyone can run it using one simple command and the results will be consistent.

Continue Reading