1. Mandatory PFD Wear: The Tasmanian Experience
Dr. L. Daniel Maxim, USCGAUX

2. Acknowledgement
Thanks to Mr. Peter Hopkins, Recreational Boating Manager, Marine and Safety Tasmania (MAST), for sharing data and insights and for reviewing a draft of this presentation
All data contained herein are for recreational boating only

3. Outline of presentation
Background
“Before and After” comparisons of fatality rates
Concluding comments

4. Background

5. Tasmania
One of the states of Australia—2008 population approximately 500,000
Capital city; Hobart
State is an island, south of Australian continent, separated from the mainland by the Bass Strait

6. Tasmania
Area of Tasmania 26,410 sq mi, which is approximately the same size as West Virginia
Coastline of Tasmania and islands ~ 3,033 miles, roughly 2.25 times that of Florida
Registered boats per 100,000 population; 5,468, similar to Florida (5,119)

7. Tasmania Tasmania has many attractive areas for boating
D’Entrecasteaux Channel, shown at left is one popular location
Other popular locations are Tasman Peninsula, Tamar River, and upper East Coast

8. Other boating areas Tasman Peninsula Boat Harbour Beach NW Tasmania
Tamar River

9. Recreational boat ownership
Results of 2007 survey shows most boats owned are ≤ 6 meters (19.7 ft) in length
All motorboats (with 4 HP and over including PWCs) account for 90% of total
Data from survey

10. Recreational boat use Boating is popular sport in Tasmania
Reported annual usage (times per year) of recreational boats in 2007 survey
Data from survey

11. Recreational Boating
Recreational boating in Australia regulated by individual states; laws differ reflecting differences in environment (risk)
Recreational boating authority Marine and Safety Tasmania (MAST) since 1997, replaced various Marine Boards around the State
MAST proactive in safety area—uses both voluntary and mandatory approaches
Useful background can be found in report at

12. Recreational boat registrations
Time series of number of registered boats in Tasmania shown at left
Still relatively few in absolute number, comparable to boats in Wyoming, but relatively rapid growth (5.8% per year)
Blue = actual data
Grey = fitted
Iteration
No Loss BETA R
D D D-01
D D D-01
D D D-01
D D D-01
D D D-01
D D D-01
D D D-01
Dependent variable is REGS
Source Sum-of-Squares df Mean-Square
Regression
Residual
Total
Mean corrected
Raw R-square (1-Residual/Total) =
Mean corrected R-square (1-Residual/Corrected) =
R(observed vs predicted) square =
Wald Confidence Interval
Parameter Estimate A.S.E. Param/ASE Lower < 95%> Upper
BETA R

13. Regulation MAST proactive:
Licenses required for all boats with 4 HP or greater—applicants being required to log 20 hours of sea time or undertake a practical course before a license is granted
Mandatory wearing of PFDs on boats under 6 meters in length (after 1 Jan 2001) when under power
Mandatory carriage of EPIRBs for these craft in coastal waters
Tasmania offers opportunity for “before and after” comparisons on mandatory PFD policies
Information on licenses and education can be found at
Q&As can be found at

14. Regulation
Focus of recreational boating safety initiatives is on the development of a “safety culture”
Regulations help to codify this culture—”it’s the way we do things here” not punitive “top down” approach
Tasmania offers opportunity for “before and after” comparisons on mandatory PFD policies—but changes in fatality rates reflect more than just mandatory PFD wear

15. Characteristics of a robust safety culture
Informed
Mindful
Reporting
Flexible
Learning
Just
The “Godfather” of safety culture is a British psychologist, James Reason. This illustration captures the essential ideas from his various publications on the relevant characteristics or elements of a robust safety culture.

16. Has regulation depressed interest in boating?
Arguably Tasmania has the most stringent regulations for recreational boating of any state in Australia
Some feared “the end of boating in Tassie”
Figure at left shows % increase in recreational boat registrations (2000 to 2009)

17. Useful publication
MAST maintains web site at
Content interesting and relevant—one publication provides summary of recreational boating initiatives

18. Data and before and after comparisons

19. Drownings Figure at left shows pre- and post-law drowning time series
Drownings relatively few and highly variable, year to year
Mandatory PFD wear for vessels under 6 meters took effect on 1 Jan 2001
  Recreational   boat drownings
Year Number
1987 6
1988 0
1989 3
1990 1
1991 5
1992 5
1993 1
1994 3
1995 2
1996 3
1997 1
1998 4
High fatalities in this year prompt review of policies
2000 3
Regulation requiring wearing of PFDs effective Jan 2001
2002 2
2003 1
2004 1
2005 0
2006 4
2007 3
2008 0
2009 1

20. Drownings
Mean drownings per year 3.5 before mandatory PFD regulation, 1.66 after; directionally consistent with hypothesis that mandatory PFDs save lives—however sample size small and series variable
2-tail “t” test shows on borderline of significance (p = assuming constant variance, p = assuming unequal variance)
95% confidence intervals on difference in means (pre- and post-) include zero
Two-sample t test on DROWNING grouped by LAW$
Group N Mean SD
Postlaw
Prelaw
Separate Variance t = df = Prob =
Difference in Means = % CI = to
Pooled Variance t = df = Prob =
Difference in Means = % CI = to
With possible outlier deleted:
Prelaw
Separate Variance t = df = Prob =
Difference in Means = % CI = to
Pooled Variance t = df = Prob =
Difference in Means = % CI = to

21. Drowning rates
Drowning rates per 100,000 boats also highly variable, but logically improved basis for comparison
Mean prelaw drowning rate 29.2 compared to 7.44 per 100,000 boats after PFDs made mandatory on small boats

22. Drowning rates
Difference in drowning rates per 100,000 boats both statistically and practically significant:
(p = with equal variance hypothesis and p = if variances unequal)
Statistical results consistent with hypothesis that mandatory PFD rule reduced annual drowning rate per 100,000 registered vessels
Two-sample t test on RATE grouped by LAW$
Group N Mean SD
Postlaw
Prelaw
Separate Variance t = df = Prob =
Difference in Means = % CI = to
Pooled Variance t = df = Prob =
Difference in Means = % CI = to
With outlier deleted:
Prelaw
Separate Variance t = df = Prob =
Difference in Means = % CI = to
Pooled Variance t = df = Prob =
Difference in Means = % CI = to

23. Box plot drownings
Box plot of drowning rates shows difference between pre- and post-law periods
Each observation corresponds to one year
Very large scatter in pre-law period reflects outlier

24. Time series analysis
Simple exponential fit (nonlinear least squares) shows that rate decreased by approximately 5.6% per year
Substantial scatter in data and possible outlier probably distorts picture
Iteration
No Loss BETA R
D D D-01
D D D-01
D D D-01
D D D-01
D D D-01
D D D-01
D D D-01
Dependent variable is RATE
Source Sum-of-Squares df Mean-Square
Regression
Residual
Total
Mean corrected
Raw R-square (1-Residual/Total) =
Mean corrected R-square (1-Residual/Corrected) =
R(observed vs predicted) square =
Wald Confidence Interval
Parameter Estimate A.S.E. Param/ASE Lower < 95%> Upper
BETA R

25. Was 1999 atypical?
The year 1999 had an unusual number of drownings, which actually spurred review of PFD policy; was this an outlier?
Probably (p < 0.05 using either Grubbs or Dixon Q test), but conclusion that drowning rates per 100,000 registered boats (pre- and post-) are statistically different holds true even if this data point is deleted
One test, the so-called Grubbs test (see calculates a 0.05 z score for this observation (among other prelaw values) of 2.83 compared to a critical value.
Another is the Dixon Q test, for which the critical 0.95 value is also exceeded.

26. What is a statistical outlier?
Some data points in a sample will be further away from the mean than what is “deemed reasonable”
Outlier points might indicate faulty or otherwise non-representative data
Statistical procedures have been developed to identify and test for outliers
It is reassuring if the presence or absence of presumed outliers has no effect on the conclusion(s), as is the case here

27. Box plot with outlier deleted
Figure shows clear differences between pre- and post-law periods
Outlier does not affect conclusion that post-law rates significantly lower than pre-law rates

28. Time series with outlier deleted
Redo of model with outlier deleted and dummy variable to capture post-law period
Best fit rate 5.3% per year, dummy variable coefficient = -5.7, right sign but NS
Scatter still substantial
Iteration
No Loss BETA BETA R
D D D D-01
D D D D-01
D D D D-01
D D D D-01
D D D D-01
Dependent variable is RATE
Source Sum-of-Squares df Mean-Square
Regression
Residual
Total
Mean corrected
Raw R-square (1-Residual/Total) =
Mean corrected R-square (1-Residual/Corrected) =
R(observed vs predicted) square =
Wald Confidence Interval
Parameter Estimate A.S.E. Param/ASE Lower < 95%> Upper
BETA
BETA R

29. Wear rates Data on lifejacket wear rates in Tasmania are not published
However, substantial anecdotal information supports claim that lifejacket wear rates in Tasmania are quite high (≥ 95%), reflecting diffusion of safety culture among boating public

30. Concluding comments

31. Concluding comments
Many studies (e.g., in Canada, the UK, Australia, and the US) have concluded that increases in lifejacket wear rates could decrease drownings substantially
Tasmania case particularly interesting because both pre- and post-law data available; though sample sizes are small, results consistent with hypothesis that lifejackets save lives

32. Concluding Comments
Safety improvements in Tasmania not solely due to requirements to wear lifejackets—rather a MAST emphasis on safety culture
Tasmania case also shows that stringent regulations do not necessarily impact participation