The price of a cyclist’s life
An interesting question was posed by Paul Clarke on Thursday on Twitter: what is the acceptable number of cyclist deaths in London per annum? I believe it was in response to cyclists calling for safety improvements following the death of a cyclist on Bishopsgate that same day.
It brought to mind a similar question I’d posed earlier: what would be an acceptable bonus for the CEO of a UK bank? In both cases, anything positive causes some degree of outcry.
But more importantly, it brought me back to an argument I’ve discussed many a time. What is the acceptable cost of safety?
Some people I speak to believe every accident is preventable and should be prevented. This, to me, is a ludicrous statement. Just as no IT system can guarantee 100% uptime, no mode of transport can guarantee that accidents will never happen.
Safety in any mode of transport can be improved. But with improvement comes cost. For many modes of transport, that cost is passed on to the customer directly.
The Boeing 747 has 0.71 crashes involving one or more deaths for every million flights (across all of its 19m flights). (As an side, the Airbus 320 range is the safest of the big players—those with over 10m flights—with only 0.10 such crashes per million flights.)
That 0.71 can be reduced. Further security checks can be introduced at airports to reduce the incidence of bombs and hijackers on board. A worldwide ban could be introduced on flying through turbulent air. The entirety of each aircraft could be checked thoroughly before each flight, and any parts showing the slightest degradation could prompt their immediate replacement.
In reducing that figure to 0.35, say, the cost of a return ticket from London to New York might increase from £400 to £4,000. A further reduction to 0.18 might increase it further to £40,000. These numbers are made up, but the order of magnitude increases are probably not far off the mark.
Those people calling for the safety improvements might cut back on their transatlantic jaunts when they hear of the associated cost hike. Indeed transatlantic flight would disappear overnight—one way of guaranteeing 100% safety, I guess.
When airlines talk of safety being of the utmost importance, they generally mean this within certain market constraints.
The cost of cycling is different. Instead of cyclists paying directly for their journeys, everyone pays for their facilities through taxation. Assuming 500,000 cyclists (there are 480,000 daily journeys, apparently), and ignoring the cost of the original road construction, the Cycle Superhighways would have cost each cyclist approximately £120. I’m guessing that they would not have been willing to pay for this, nor would they be willing to pay directly to implement further safety improvements.
If it costs more per death saved than it would cost the NHS to save a life, should the money be diverted instead to the NHS? (A reminder of the trolley problem: should you actively sacrifice someone’s life if you know it will save five other people’s lives?)
Sixteen cyclists were killed on London’s roads in 2011. The highest such figure was 33 in 1989, the lowest: eight in 2004. What is an acceptable number? And what is the acceptable cost of achieving that?
Confounding compound interest
As a kid, I never understood why a seemingly arbitrary and unwritten rule stipulated that compound interest would apply daily. That is, if you were asked to calculate compound interest over the course of a year, the interest accrued would become part of the principal sum at the end of each day. I never understood why it shouldn’t apply every hour, minute or second.
Below is the effect of differing frequencies with which the principal sum is topped up, each on an initial principal sum of 10,000 and a compound interest rate of 20%. The figures show the resulting sum after one year of investment.
- Annually (i.e. simple): 12,000
- Quarterly: 12,155.06
- Monthly: 12,193.91
- Weekly: 12,209.36
- Daily: 12,213.36
- Hourly: 12,214.00
- Every minute: 12,214.03
- Every second: 12,214.03.
I guess for the sake of threepence, I needn’t have worried.
Is John Lewis’s TV advert driving its record sales?
On Monday, Neil Mortensen, Research & Planning Director at Thinkbox, tweeted the following:
Record sales for John Lewis last week, biggest week ever. #tvworks
This drew a wry retweet from John Willshire of Smithery.
The same must be true for all last week’s ads then?
It’s true. Claiming success of TV advertising for a record sales week two weeks before Christmas is somewhat glib. Imagine the following fictitious tweet from Standard Fireworks on 6 November.
Record sales for Standard last week. Biggest week of the year so far. #tvworks
While the evidence may indeed be there for TV advertising driving John Lewis’s record sales, the implication was that their high sales were wholly a result of the much-lauded TV ad. I expect that most companies tailoring their offerings to the Christmas market have seen increased, indeed record, sales over the last few weeks.
Stripping out seasonality is important, as is stripping out the effect of other media. Otherwise, statements such as these either artificially elevate the impact of TV advertising; or else are treated with the important pinch of salt that they warrant.
xkcd and osirra’s Christmas music analysis: the difference a Pond makes
At 5am this morning, the latest xkcd blogpost was published, titled Tradition. (xkcd remains my favourite blog on the interweb, btw.)
I read it at around 0945. And by 1015 I’d posted the British equivalent. (The chart was built in Excel using an xkcd-esque font that I downloaded for the very job.)
The original post showed the 20 most-played Christmas songs (2000–2009 radio airplay) by decade of popular release. Nine of the entries fell in the 1950s and a further seven fell in the 1940s. Of the remaining four, two were in the ’60s, one in the ’70s and one in the ’30s. The most recent song in the chart, Feliz Navidad, was written in 1970, 41 years ago.
The post linked the surge in music with the postwar baby boom, suggesting that Christmases since that boom have merely tried to recreate the Christmases of that time.
I was certain that the UK equivalent chart would look wildly different, so I set out to show this. And wildly different it is. (Click the chart for a bigger version.)
I was expecting the median decade to be the ’70s, but in fact eight of the songs were released in the ’80s, a further six in the ’70s. The earliest charting number was Bing’s White Christmas from 1942, the most recent being the 1994 offerings from Mariah Carey (second best Christmas tune of all time: fact) and East 17.
The UK’s 20 songs are all far more recent, with a mean release year of 1977. (America’s 20 most-played average to 1951, each song being on average over a quarter of a century older than its UK equivalent.)
There is a strong musical tradition at Christmas in the UK, one that I unreservedly love. The Christmas Number One is a big deal (or at least it was until X Factor made it so much more formulaic), and I get the impression that modern Christmas music is far more accepted and expected in the UK than it is in the US.
A lovely piece of analysis by xkcd of the US market. But, if I may say so, a much more lovable piece of analysis by myself of the UK equivalent. Without xkcd, I wouldn’t have thought to do it. But without the wonderful British Christmas music market, I wouldn’t have been able to paint such a fabulous picture.
A short analysis of £48.8bn of government contracts
My friend Alan today pointed me to this page on data.gov.uk. It gives an estimated pipeline of the big tenders that are expected to take place within this Parliament, including those over £5m in value. (There are actually 12 tenders included valued under that amount.)
Of the 208 initiatives, 203 have a maximal contract value, the average being £240m. This is heavily skewed by DWP’s facilities management contract (a maximal value of a cool £9.542bn) and HMRC’s Aspire contract for ICT provision (a mere £8.5bn). Both of these are being let in 2017. Yikes! (The next largest is the provision by the Home Office of Tetra radio services to police forces across the country,at £3.8bn.) From this albeit top-heavy sample, the top 0.5% of projects (one) make up 20% of the revenue.
The lowest value tender on the list is for DH’s website services: £1.4m. Although you can pick up courier services for the Government Procurement Service itself for £2m.
If you take the three biggest projects out of the mix, then we have the following headlines:
- 61% of the projects fall within ICT, accounting for 65% of the maximal spend. A further 25% fall within facilities management, accounting for 22.6% of the maximal spend
- Energy projects have the highest average maximal spend, at £222m. Property projects have the lowest average, at £47.5m
- The Government Procurement Service has the highest total spend, at £7.2bn, followed by the Ministry of Defence (£4.9bn), the Department for Work & Pensions (£4.1bn) and the Department for Transport (£3.0bn)
- If we look instead at averages, then even stripping out its FM contract as an outlier, the DWP tops the list with an average contract value of £690m across its six entries. This is more than double that of the next highest: the MoD at £326m across its 15 projects
An analysis of the 7,466 Premier League games
Between the 1992/3 season when the Premiership began and the end of the 2010/11 season, the biggest goal difference in the Manchester derby was five, United beating City 5–0 in the 1994/5 season. The biggest away victory in the fixture was United’s 3–0 win that same season. Not a good year if you were a City fan. The former record was equalled today, with City’s 6–1 drubbing of United at Old Trafford; the latter was shattered.
Across all 7,466 Premiership/Premier League games, only 2.3% have seen seven or more goals. (Portsmouth’s 7–4 win over Reading in 2007/8 holds the single game goal record, btw.)
If we ignore the home/away victor, 1–0 has been the most common scoreline (18.52% of games), followed by 2–1 (14.91%), 2–0 (12.63%), 1–1 (12.22%) and 0–0 (8.68%).
Two is the most likely number of goals (24.85% of games seeing this many goals), followed by three (20.80%), one (18.52%), four (14.26%) and zero (8.68%).
If you want to see lots of goals from your team while they’re at home, buy a Man Utd season ticket, with an average of 2.21 goals scored, 4.34 goals per season more than their closest rivals, Arsenal. (Steer clear of Wigan Athletic, with 1.06 goals scored by themselves per home game.) If you want to follow a team around the country, choose Man Utd, with 1.74 away goals per game. Of the current Premier League crop, don’t follow Norwich around, at 0.68 away goals per game.
If you just want to see goals, irrespective of who scores them, follow QPR, with 3.01 goals seen per game. Steer clear of Stoke City, with only 2.30.
Fabulous dataset, btw. Took a little compiling.
How to report exam pass percentage increases
When the GCSE and A-level results came out back in August, I got to thinking about how increases in percentages should be measured. There was the usual mathematical heathenry, with people talking of percentage increases when they meant percentage point increases. But that’s another story.
Even when percentage points are reported responsibly, they’re open to misinterpretation. Not all five point increases are as easy to achieve. If 25% of grades are an A, then it’s arguably easier to achieve a five point increase than if your current percentage is 90%.
And even percentage increases can be confusing. A 20% increase on the 25% takes you to 25%. But a 20% increase is not possible on the 90%.
Maybe comparisons would be best achieved if increases were reported as the percentage of fails that are now successes, or in other words, penetration into the failures. So an increase from 20% passes to 25% would be reported as 6.25% penetration into the fails, as would an increase from 90% to 90.625%. Arguably, each is equally difficult to achieve.
Or else, maybe it should be reported as the percentage penetration into the failures, divided by the prior pass rate. So 20% to 25% would be reported as a 31.25% weighted penetration (stop it), as would an increase from 90% to 92.8%. This latter coefficient takes into account the difficulty of getting people to pass. At 90%, it’s arguably easier to convert the failures than if the pass rate was at 20%.
Although while mathematically these quotients might be a better representation of the truth, explaining them to the masses might be tricky.
This is what I think about of a Friday night.
The British weather: stop fucking whinging
I am royally fucked off with people whinging about the quality of the British summers. So I did some analysis.
I analysed the mean temperature at the Hadley Centre in Exeter. Every day from 1 January 1772 through 30 September 2011. 87,566 data points in total. I removed 29 February data to allow easier cross-year comparisons. And here are the results.
Thus far this year, the 2011 average daily temperature has been higher than the 238-year average on 152 (55.7%) of the 273 days of the year so far. Where the temperature was higher, it was higher by an average of 2.72C. Where it was lower, it was so by 1.40C. And so across all 273 days, the average daily temperature in 2011 was, on average, 0.95C higher than the historic average for that day.
If we limit the comparisons to the last 50 years, it’s much closer. Temperatures on 137 days of the days thus far have exceeded the average, those on the other 136 have been lower. But even taking that into account, the average temperature across all days has exceeded the average by an average of 0.58C.
If we limit analysis to summer months (which I’ve classified as April through September), then against the average we have been hotter this year on 50.8% of days (47.0% of those in the last 50 years). But across all days, our average daily temperature this year has exceeded the day’s average by an average of 0.73C (0.50C in the last 50 years).
So please. Stop fucking whinging.
Much credit to the Met Office for its lush dataset.
Putting September 11 into perspective
In the four attacks on 11 September 2001, 2,977 victims died. (A further 19 deaths were of the hijackers themselves.)
In the US, 2,977 people are murdered every two and a half months. The same number of people die every 26 days in car crashes.
The relatives of these people don’t have a memorial or an annual news furore associated with the death of their loved ones.
11 September 2011 was unique. It was a crime of enormous proportions, in which a huge number of people died in a unique circumstances, ones with enormous political ramifications. But the cumulative loss and direct heartache is no greater, no less, than that associated with the events detailed above.
This post in no way intends to denigrate those lost on 9/11. It merely intends to put it into perspective.
GCSE results: edging towards perfection
It seems that historic GCSE attainment data is hard to access. Understandably so. The government, and its predecessors for the last 29 years, hardly have a good story to tell.
Neither the ONS, nor DfE nor data.gov.uk seems able to furnish me with historic attainment figures. Specifically, I’d like the percentage of pupils attaining five or more A*–C grades by year. And I’d like to know the grade distribution by subject (and overall) by year.
I’ve managed to cobble together a dataset containing 2001/2 data and 2006/7 data showing the proportion of A*–C grades by primary subject by geography and gender. It seems that subsequent years’ data isn’t available. For the sake of brevity, let’s wrongly refer to an A*–C grade as a pass.
Over that five-year period, the pass rate in core subjects has gone up by 8.9 percentage points, from 31.6% to 40.5%. Assuming linearity, everyone should be passing these subjects by 2039.
Over the same period, Maths passes have risen by 4.7 percentage points, from 49.2% to 53.9%. And English by 4.2 points, from 56.8 to 61.0%.
So as a nation, it’s clear that we’re becoming more intelligent. Or are we?
Of course we’re not. The whole thing is a farce. The overall pass rate (in the ridiculous “turn up and you pass” sense of the word) has risen for the 23rd consecutive year, now sitting at a heady 98.7%. The qualification has become meaningless. A grades were the first to become worthless, A* grades being introduced in 1994, soon becoming the new currency of choice. And now in certain circles, anything less than ten results containing the letter A is seen as failure. How on earth employers or colleges evaluate the relative merit of candidates I have no idea.
As I’ve said before, there needs to be some normalisation. We are not getting more intelligent as a nation. The percentage of A* grades awarded each year should not change. Ever. There should be a forced curve for each subject. It would give predictability to those organisations interested in those pupils. But perhaps more importantly, it would rid the world of trite, vacuous news stories every August, accompanied by attractive leaping girls.
This won’t happen, of course. Instead, the youth of tomorrow will edge towards perfection, everyone becoming equal and indistinguishable.