Sunday, September 25, 2005

Roadkill myths II: Israel is worse than any civilized country

In this second installment of my series on road accidents in Israel, I promised to address international comparisons. This brings me straight to the second prevalent myth.


Myth II: Israel is worse than any other civilized country

This is a bit trickier to tackle than the first myth, since international comparisons are a tricky thing. For example, different countries have different definitions of what it means to be a traffic fatality. How soon after the accident does a victim have to die to be considered a road death?

Also, different countries gather and report different pieces of data, and use different methodologies in collecting them. Who gathers the statistics and how reliable are they? When data must be estimated, on what basis is the estimate made?

Finally, countries vary greatly in their road infrastructures, in vehicle use, in general levels of development and prosperity. What constitutes the typical "civilized country" Israel should be compared with? The wealthy U.S., with its high levels of vehicle ownership and low fuel prices? Other countries with Israel's level of economic development? Countries where most driving is on empty rural routes or in jammed urban areas?

Keep such factors in mind when evaluating the data. You can't draw meaningful conclusions from narrow statistical differences. That doesn't mean you can't assess the general trends, though. Let's see what we can do.


Per capita first
As before, there are (at least!) two main ways to compare fatality rates: per capita or per distance driven. As before, we'll look at both.

Per capita rates tell you how likely an average individual is to be killed in a road accident in a given year. In a sense, it answers the question, "How dangerous is it to live in this country?" with regards to traffic accidents.

Here's the answer for 2003:


Data is from this source; note the comments at the bottom of the page and remember the limitations of this sort of data. I've left out very small countries due to volatility of data. Incorrect Israel data point in this source has been corrected; Israel data for all graphs is from the Central Bureau of Statistics.

For easy identification, I've shown Israel's data in green, and a handful of major western countries in orange. Of the 45 countries shown, Israel ranks 6th. That is, only five countries had a lower fatality rate per capita. The United Kingdom placed 2nd, Japan 7th, Germany 13th, Canada 16th, France 18th, Italy 23rd, the United States a lowly 35th, and South Africa 44th between Russia and Malaysia.

In fact the U.S. fatality rate was more than double Israel's. The average person is twice as likely to be killed on the roads in the U.S. in any given year than in Israel. Israel's fatality rate is 16 percent higher than that of the supersafe U.K., though.

Surprised? Doesn't everyone know the roads are dangerous in Israel? Isn't it obvious to any American visitor, for example, that drivers are less cautious, cars are not as solid, and the roads are not as well maintained as they are at home?

All of that may be true, but that relates to how much risk is involved in driving a particular stretch of road. That is, as I explained in the previous installment, it's about not fatalities per capita but fatalities per distance driven.


Now per kilometer
Let's look at that next. The number of fatalities per distance driven answers the question, "How many people are killed in an average-distance trip in this country?" Again for 2003, the fatality rate per billion vehicle kilometers (fewer countries report this data, yielding a smaller but no less enlightening graph):


Data is from this source. Note that overall nationwide kilometers driven can be difficult to estimate, and methods may vary among countries.

By this measure, Israel fares less well. Out of 23 countries reporting, Israel places 15th. On the other hand, we're in good company. France, Ireland, Japan, Austria and New Zealand are within 10 percent of us. Israel's fatality rate is 23 percent higher than that of the U.S. but 21 percent lower than that of Belgium. Compared with the top-ranked U.K., we're 52% worse.

As you expected, the roads in Israel are, on the average, not as safe as those in the U.S. But the gap is not enormous. Clearly there's plenty of room for improvement. Still, we are far from having the worst road safety record of any developed country.

If Israel's roads are more dangerous than America's, why are so many more Americans killed per capita than Israelis? Simple: Americans drive much more than Israelis - over two and a half times as much on average. Cars in Israel are much more expensive, as is fuel, and the average salary is significantly lower. So the average Israeli spends much less time on the road than the average American, and thus his exposure to the risks is much lower.

From the above statistics, we can compute the average distance driven per capita in each country:



All else being equal, the less people drive, the fewer accidents there will be!


As time goes by
I mentioned the need for improvement. As we saw last time, road safety in Israel is improving over the years. But other countries aren't standing still. How does Israel's record stand up over time?

Again, let's start with the per capita data:


Data is from here.Again, incorrect Israel 2003 data point in this source has been corrected.

Sorry about the messy graph, but the overall picture is clear enough. Per capita fatalities are improving in most countries, some faster than others. Many, you can see, have improved faster than Israel (in green); on the other hand, most had a much worse starting point. Israel was one of the countries with low per-capita fatality rates in 1988; it remains so today. In the interim, though, the gap has narrowed.

Let's simplify that data by looking at just the percentage improvement:



Israel's fatality rate has dropped by over 40% in fifteen years, while other countries in the middle of the pack also improved by 35-40%. The biggest improvements were over 50%, yet all of the countries with massive improvements were the ones with the worst starting points - they had the most room to improve.


To round out our picture, let's see how per-distance fatality rates have changed over the years:


Data is estimated from the graphs in this PDF file; I haven't been able to find the raw statistics. View the file to see the yearly improvements by country.

As the data show, Israel's safety record was poor back in 1980, about twice as bad as developed countries such as the U.S., the U.K. and Scandinavia. By 2003, though, as we saw above, Israel was only 52% worse than the U.K. Note also that all the countries in the sample had safer roads in 2003 than even the best of them had in 1980. This demonstrates the power of gradual improvements over time.


For a better focus on Israel's improvement over time, let's plot that on its own:



Israel is among the leaders here, with an improvement of over 70%, more than the U.K. As with the per-capita improvements above, the countries which improved the most (including Israel) were generally the ones with the worst starting points.


It's all about economics
How could Israel show one of the biggest improvements in per-distance fatalities, but place below average in improvement in per-capita fatalities? I don't have the statistics, but there's only one possible explanation: Compared to other countries, Israel has had a far greater increase in road use per capita. Israelis drive much more today than they did in 1980, and that has kept the per-capita fatality rate from falling nearly as fast as the per-kilometer rate.

As long as Israeli economic growth continues at a strong pace, this will continue to be the case. Economic growth has two effects: Society can afford more road safety, but individuals can also afford to drive more. As the roads get safer, the public's exposure to the risks of the road rise, moderating the impact of the safety improvements. I demonstrated this point in the previous posting, but the international comparison highlights it again.

To summarize: Israel's per-capita road fatality rate is lower than average for developed countries, while its per-kilometer fatality rate is higher than average. The discrepancy is due largely to Israel's low rate of road use. It's safer, on average, to drive a given distance in the U.S. than in Israel, but people do it far more often.

Most importantly, over time Israel, like other countries, is making substantial improvements in road safety, gradually closing the gap in per-distance fatalities. It would be great to improve even faster, but Israel's record is nothing to be ashamed of.

Which brings me to one last point, and one last graph. Note that most of the other countries in these comparisons are much wealthier than Israel per capita. Like everything else in the physical world, safety costs money: Better roads, better cars, better driver education, better enforcement. All else being equal, rich countries can afford more road safety than poor ones. We can see this by ranking countries by GDP per capita:


GDP figures from here.

Of the countries near Israel's level of economic development, Israel has the lowest rate of fatalities per kilometer driven. In fact, we're well within the range of road safety parameters achieved by some far wealthier countries, such as France, Austria, Japan, Ireland and Belgium. While there's still plenty of room for improvement, I'd say Israel makes a pretty good showing. By no means are we "worse than any other civilized country"!


I'm still thinking about what issues to address in future installments in this series. Suggestions are welcome.


Update (Sep. 26): In a comment, David Boxenhorn of Rishon Rishon asked for a graph related to the last one. I've taken the liberty of answering with what he really meant, not what he asked for. The question is: How safe are the roads in different countries relative to their GDPs? That is, if richer countries can afford more road safety, how well are countries doing considering how much they can afford?

"The amount of road safety" is the inverse of fatalities per kilometer: the average number of kilometers driven per fatality. The higher the figure, the safer the roads.

David asked for one graph; I'll give him two. First, a scatter plot of kilometers per fatality versus GDP per capita, along with a linear regression estimating the relationship between them. (I haven't tried any sophisticated econometric analysis; this is the standard regression function built in to my spreadsheet.)



The distance of each dot from the line indicates how much better or worse the country's safety record is relative to what one would expect based on its GDP per capita. The United States, Belgium, South Korea and Greece, for example, are significantly below the line, indicating a poor level of road safety relative to GDP. Significantly above the line are Israel, the UK, Finland, the Netherlands and Australia, among others, indicating a low fatality rate relative to national income.

To summarize this variation in a single statistic, we can divide the amount of road safety by GDP per capita. I've decided to call this the Bang for a Buck Index: Kilometers per fatality divided by GDP per capita. The higher the index, the more road safety "bang for the buck":



Israel has nothing to be ashamed of. Not at all.

16 comments:

David Boxenhorn said...

I would like to see one more graph: (fatalities per mile driven) / (GDP per capita).

Cosmic X said...

ZB,

Thanks for this excellent and informative post.

Zman Biur said...

David,

How about two? Reload the posting; I've addressed your implied question in an update.


Cosmic,

Glad you liked it.

Jameel @ The Muqata said...

Do your stats include all the accidents in the West Bank as well?

Zman Biur said...

Jameel,

Good question. The answer is no.

Ezzie said...

Excellent post! When you say it doesn't include the West Bank, does that mean it doesn't include the people in the total population either? And what about the Gush Etzion/Neve Yaakov/Maale Adumim areas?

Zman Biur said...

Ezzie,

Thanks.

I said international comparisons are tricky; sometimes deciding how to handle data in a single country is tricky too.

Regarding the "West Bank": I'm using the official data published by Israel's Central Bureau of Statistics. I'm pretty sure that the road accident data does not include the West Bank and Gaza, but I believe the Israeli population data does include Israeli citizens resident in the settlements. I'm not sure whether it includes non-citizen Arabs resident in eastern Jerusalem.

(Specfically, Neve Yaakov is part of Jerusalem and included; Gush Etzion and Maale Adumim are not.)

While that might be somewhat inconsistent, I believe it is the best way to handle the data. Counting accidents in the West Bank would require expanding the dataset to include the Palestinians. Aside from being culturally, economically, developmentally and geographically very different from Israel - thus making it meaningless to consolidate their accident data with Israel's - the Palestinians collect their own road accident data which is not reported to Israel. It would also become impossible to draw conclusions regarding Israel alone.

On the other hand, many if not most Jewish residents of the West Bank travel frequently on the roads in "Israel proper", thus contributing to its accident rate, so it's not entirely unreasonable to include them in the population.

In any case, they constitute at most 3.5% of Israel's population, so whether or not they are included would have a marginal impact on the results. This is just one of many factors which introduce uncertainty and margins of error to international comparisons. Israel's total population grows at 2% a year - do we use the population at the start, end or middle of the year? Do we count tourists and foreign workers? Etc. (We could ask similar questions about other countries.)

Frankly, differences of less than 10% across countries are probably meaningless, maybe even 20% or more. There are too many uncontrolled variables.

David Boxenhorn said...

Thanks, Zman!

I think I "really" meant what I asked for, but what you gave is just as good (okay, better...)!

Zman Biur said...

David,

Trust me - the statistic you asked for isn't very interesting. I checked.

I'm not quite sure how to interpret it, either.

David Boxenhorn said...

It's just your chart #2 normalized for per capita GDP. You just have to get the scale right to make it look like something (I think). But I like your graphs just fine, and I agree that bang-for-buck is a good description. The only problem is that people think in terms of reducing the fatality rate. Increasing the number of miles driven per fatality is a very unnatural way for most people to think, I think. Also having fatalities in the numerator allows for direct comparisons between ways of measuring it, e.g.: According to measurement A Israel is X, according to B is Y, etc.

BTW the statistic that they give in the papers to make Israel look bad is fatalities per km of road - which is totally ridiculous. It just shows that Israel has crowded roads.

Zman Biur said...

David,

The problem is that the relationship between the fatality rate and GDP is both inverse and nonlinear. The plot is in
this file
.


Since road safety increases, more or less linearly, with GDP, it makes sense to divide an index of road safety by GDP, yielding a measure for how much safety one achieves per unit of GDP.

The fatality rate, however, decreases, apparently asymptotically, with increases in GDP. Dividing by GDP yields an index of "fatalities per kilometer per unit of GDP", which effectively just lowers the fatality rate even further for rich countries and raises it for poor countries. It doesn't measure how much you get "per GDP", since you (hopefully) get fewer fatalities with higher GDP.

Here's the graph. How would you interpret it?




You're right that fatalities per quantity of road is absurd - if Israel built more roads and closed them off to traffic, it would place higher!

Fatalities per number of vehicles registered is also meaningless. Cars which sit in driveways naturally don't contribute to the accident rate.

In both cases, Israel has a much lower density of roads and rate of vehicle ownership than other developed countries, so such measures would make Israel look exaggeratedly bad.

Zman Biur said...

Clarification: When I said "lower density of roads", I meant "less road surface per population", not "fewer vehicles per amount of road surface"!

Zman Biur said...

I've doublechecked the policies regarding Judea, Samaria and Gaza. The population statistics do include Jewish residents of the settlements. The accident statistics include the (few) accidents which take place within the settlements themselves, but not on the intercity roads in the territories. A separate appendix is available regarding accidents in the territories which are attended by the Israeli Police, but the data used here does not reflect them.

David Boxenhorn said...

Oops! You're right. I guess I wanted (fatalities per km) * (GDP per capita), which would be the inverse of bang-for-buck, i.e. bucks-for-bang (buck-bangs per km?).

It would be nice to be able to compare various measurements. You could normalize all of them by setting Israel=100.

Jack said...

Interesting.

Miriam L said...

I'm definitely not driving in Slovakia.