Review of the Victorian Wild Dog Management Program and Recommendations for Future Approaches

Noting the statistical limitations to our analysis of trends in baiting levels, our stakeholder conversations indicate that the increase in baiting levels over the most recent three years 2012/13 to 2014/15 and improvement in effectiveness for baiting from 2013/14 to 2015/16 (Figure ) has resulted from a range of factors, including improved strategic targeting of wild dog control measures. Although reactive trapping control continues in response to incident reports, the timing and location of proactive baiting and trapping has become more targeted and more importantly better communicated with public and private control activities more aligned and coordinated.

Another critical component of improved delivery has been improvements to the workforce undertaking the public control measures. These changes include:

increasing the use of casual staff to increase the capacity to contribute towards increases in seasonal activity and targeted local activity;

improving workplace arrangements by enabling more use of proactive control measures and improving transparency of operations in a more timely and efficient manner; and

improving capabilities through training in new control technologies and improving communication methods and skills.

This includes positive program changes such as:

increasing and shifting the areas of baiting to better target the source of wild dogs;

seasonal planning of community baiting programs to align the timing of baiting to better reflect the effect of seasonal conditions on wild dog risk, breeding cycles and bait lifespan; and

improving the timing and duration of monthly baiting cycles to better reflect community needs.

Key mechanisms to lever this shift have been:

zone management planning;

community consultation on the plans and public control activities; and

improved private access to baits (particularly fresh meat baits) including accreditation, training and bait provision.

The Baiting Coordinator Project (BCP) has been a key project delivery vehicle for these mechanisms. The BCP estimates that over the life of the program:

there were 17 community baiting groups;

147 landholders participated in the baiting programs covering 120,000 hectares; and

participants laid 60,000 baits in areas where previously there had been little or no community baiting on private land.

Since its inception in 2012, the BCP has enabled;

119 members to complete AgVet Course (Agricultural Chemical User Permit) training;

147 members to receive 1080 endorsement;

95 members to receive trap training; and

GIS mapping products to be provided to better capture and record community control efforts.

We note that while the program has shifted to a state-wide focus, progress on monitoring the initiatives is more advanced in some control regions than others. In particular there remains a substantial gap in some aspects of program delivery such as recording and monitoring program activities in the North West of the State.

While there have been anecdotal improvements in the wild dog controller work-force, this is an area for ongoing focus given the importance of interaction between wild dog controllers and landholders. This includes continuing to build trust relationships with landholders, improving proactive communication and negotiation, and improving transparency of workplace activities and performance.

Improvements have also occurred in the effort directed at broader community relationship management and coordination. As noted earlier, a major improvement in the effectiveness of wild dog control lies in the increased focus of operational staff on community engagement, together with the appointment of the Community Engagement Manager. The Australian Wool Innovation (AWI) staff have played a major role in improving community engagement and communication. Across the board, many stakeholders indicated that these appointments have had a profound effect on organisational and general program performance by improving program transparency and capacity to negotiate, adjust and target public and private effort.

3.3.3Overall effectiveness of inputs in producing outcomes

Livestock deaths and injuries from wild dogs (note most are unconfirmed causes) fell over the period 2012/13 to 2014/15 (Figure ) yet there was a similar number of dog attacks over the same period (Box 3). What this means is unclear as the timeframe is short and there are factors other than wild dog control measures that can influence wild dog attacks (e.g. rainfall, wild dog breeding conditions).

The effectiveness of the two key control measures (ground baits and traps) in producing outcomes (i.e. dogs destroyed and reduced livestock losses) has improved over the last three years – mostly from improvements from 2013/14 to 2014/15. Moreover, there has been a decrease in the number of confirmed livestock deaths per bait/trap night between 2013/14 and 2014/15 (Figure ). However, we believe that this improvement should be interpreted cautiously due to the short time period analysed and the known seasonality issues.

Additionally, we undertook statistical analysis (e.g. regression analysis) to examine the relationship between confirmed livestock deaths and the two key input variables (baits laid and traps checked). This analysis was undertaken using regression analysis². The analysis was undertaken for the aggregate Victorian data set and also for some of the regions. The regression analysis was undertaken in various ways including:

Single variable equations (e.g. livestock deaths as the dependent variable and baits laid as the independent variable);

Multivariate equations (e.g. livestock deaths as the dependent variable and baits laid and traps checked as the independent variables);

Incorporating a one season lag in the analysis since baiting and trapping typically will have an effect sometime in the future on livestock deaths and current levels of DEWLP trapping and baiting are often reflective of very recent livestock death incidents;

Incorporating seasonality via a dummy variable where appropriate where community baiting is undertaken.

This analysis did not reveal any logical statistically significant relationships at either an aggregate level or regional level. Moreover, while some of the relationships were statistically significant at the aggregate level, this was not considered a logical outcome because:

In the cases in which there were statistically significant relationships the analysis indicated that higher levels of baiting or trapping was leading to higher levels of livestock deaths. This is illogical since control measures are not likely to increase livestock deaths. Rather, the result most likely reflects randomness in attack patterns and the fact control activities increase in response to increased attacks and there can be a long time (perhaps beyond the timeframe of the dataset) before clear long term decrease in activity occurs.

In many cases the regression analysis at the regional level produced no statistically significant relationships and where there were statistically significant relationships the results were, again, counterintuitive.

It is likely that the same analysis would benefit from a longer time series and there is a benefit in undertaking similar analysis at some time in the future. It is also likely that there are other factors not included in the regression modelling that are impacting the results. For example, the effectiveness of baits is impacted by wet weather in localised areas.

Figure : Livestock deaths from wild dogs – overall

Box 3: Performance against strategic objectives that are program output related Strategic objective 2: After 3 years (2015) the number of reported dog attacks each year is 15% less than the number in previous years Incidents of dog attacks have been recorded since 2012/13 including details about the location and livestock involved in the attacks. This information is shown below in Figure . Figure : Dog attack reports – overall There has been a slight decrease in the number of reported dog attacks between 2012/13 (607 reported attacks) to 2014/15 (591 reported attacks). This equates to a 3% drop across the two year period. The number of attacks in 2013/14 was 807. We note our previous comments on drawing conclusions from these movements.

Figure : Effectiveness of inputs in producing outcomes – overall



Notes: It is assumed that each bait laid is checked after 26.8 nights on average and a wild dog controller has, on average, 15 traps in place in a typical working week. Also note that in the graph it is assumed that a bait does not deteriorate over the 26.8 nights such that it becomes ineffective (e.g. due to weather).

3.4Regional program performance

3.4.1Regional zones for comparison

Four regional clusters were developed in consultation with DEDJTR and DEWLP to enable the comparison of high level program data (Figure ). This involved grouping the current fourteen zones into four clusters to account for commonality in scale, topography, and program activities and farming over lay within zones. This allows for a more appropriate comparison across Victoria since the four zones were constructed taking into consideration that wild dogs will move across the current fourteen zones and control measures in one zone may impact outcomes in another zone that is close by.

Note that our assessment of statistical significance in the regional program performance section of the report considered t-stats at the 95 per cent confidence level (2-tailed).

Figure : Regional management zones and clusters

3.4.2Regional program inputs: control measures

The overall level of baiting (community and DELWP) is seasonal in nature with higher levels in autumn and spring (Figure ). Community baiting is clearly seasonal, occurring in Spring and Autumn (Figure ).

In contrast, the timing of DELWP public baiting is much less seasonal (Figure ) with Regional Cluster 1 exhibiting a statistically significant downward trend in DELWP baiting levels – all others clusters do not have statistically significant trend.

Figure : Overall baits laid – by region



Notes: includes DELWP ground baiting, community ground baiting and aerial baiting.

Figure : Community baiting – by region

Figure : DELWP ground baits laid – by region

Trapping does not follow a seasonal pattern (Figure ). The lack of seasonality reflects the fact that the overwhelming majority of trapping is reactive and occurring continually throughout the year in response to landholder reports of attacks, and that only a very small number of traps are set as a proactive measure.

Noticeably, there is no statistical significant trend upwards or downwards in any of the clusters for traps checked.

Figure : Traps checked – by region

A number of factors may explain the differences in the seasonal approach to trapping and baiting:

work practices determining a baseline of control. In each region, controllers undertake trapping as the core labour activity. The work undertaken by a wild dog controller in a zone in any given week is determined by the requirement on them to be 100 per cent compliant with the regulations, the associated practicalities of checking each trap at no more than 72 hours after setting, and that work is only undertaken during the week (Monday to Friday). All wild dog controllers are allocated an annual target for baiting that must be completed;

reported attacks are not highly seasonal and as a result reactive control is not seasonal (Figure );

the number of traps deployed is largely a function of reported dog attacks since traps are typically deployed in response to reports by property owners of attacks; and

the close proximity of control activities between zones in the North East — control activities in the Mitta Mitta and Ovens and Alexandra and Mansfield zones are in close proximity to one another and control activities within one zone are considered by the program to impact on neighbouring zones.

3.4.3Regional effectiveness of inputs in producing outputs

In terms of regional clusters, there is a statistically significant upward trend over time in the effectiveness of baits in Regional Cluster 1. This is not the case in other regional clusters. Figure illustrates the change in effectiveness in each of the clusters over time. One of the issues in examining trends is the short time period from which to draw solid conclusions.

The effectiveness of baits has been measured by the ratio of DELWP ground baits taken by wild dogs to baits laid.

Figure : Ground baits taken by wild dogs per baits laid – by region

Figure : Dogs trapped per traps checked – by region

3.4.4Regional effectiveness of inputs in producing outcomes

The number of confirmed livestock killed and dog attacks by regional cluster over the period are shown in Figure and Figure respectively. Additionally, Figure and Figure illustrate the effectiveness of inputs (DELWP ground baiting and trapping) in producing outcomes (less livestock deaths and/or reported dog attacks).

For each of the four regional clusters there is no statistically significant trend upwards or downwards³ in the numbers of confirmed livestock killed and dog attacks. However, Regional Cluster 2 is statistically significant in terms of the effectiveness of inputs in producing outcomes for both confirmed livestock deaths per 10,000 bait/trap nights and reported dog attacks per 10,000 bait/trap nights). This is not the case for other regional clusters.

The effectiveness of inputs in producing outcomes illustrates how effective the control strategy could be in delivering desired outcomes. For example, a positive outcome is for less inputs to be required to deliver a similar outcome over time. In other words, there is the same amount of livestock deaths or dog attacks with less baiting and trapping effort. A positive outcome could also be less livestock deaths or dog attacks with the same baiting and trapping effort.

Therefore, there is a positive trend in the effectiveness of inputs producing outcomes in Regional Cluster 2 as less livestock are being killed for the same amount of bait/trap nights.

Figure : Livestock deaths – by region

Figure : Dog attack reports – by region

Figure : Effectiveness of inputs in producing outcomes – livestock deaths by region

Figure : Effectiveness of inputs in producing outcomes – reported dog attacks by region

3.5Optimising control activities

3.5.1Optimising mix of ground baiting and trapping

Baiting and trapping have different attributes that impacts their relative usefulness as a control measure. Overall, the cost per dog killed using baits is almost one-third that of using traps ($1,300 compared to $4,095) – (Table ). There are two offsetting factors that generate this result. First, the cost per night is lower for baits than traps. Secondly, traps kill more dogs per bait/trap night.

The estimation of these comparative measures is contained in the following sections.

Table : Comparative measures modelled to indicate – comparative ground baiting versus trapping costs

3.5.2Comparing the cost of ground baits versus traps

In this section we consider the relative cost and impact of the two key control measures – ground baiting and trapping. The two measures are compared with reference to the cost per night for one unit of bait or trap in order to ensure comparability between the measures.

We have estimated ground baiting and trapping costs using a bottom up method and then used a top down method to ensure that the bottom up costing is reasonable compared to total wild dog costs.

Using this approach, ground baiting costs per night are estimated to be lower than trapping costs per night (Figure ). Furthermore, trapping costs are estimated to be almost 30 times higher than baiting costs per night.

Figure : Comparative cost per bait/trap night

The composition of these costs is illustrated in Table .

Table : Composition of costs

Control type	Baiting	Trapping
Labour cost per night	$0.77	$23.40
Bait/trap purchase cost per night	$0.11	$0.31
Travel cost per night	$0.09	$4.75
Total cost per night	$0.98	$28.46

Source: MJA analysis

The cost types are:

labour cost of staff (wild dog controllers) either placing baits in the ground or setting traps;

bait/trap purchase cost, noting that traps are often reused; and

travel cost relating to travelling to different locations to place baits or set traps.

The key assumptions for each of these cost types for baiting and trapping is contained in Table and Table respectively.

Table : Baiting cost assumptions

Control type	Key assumptions
Labour cost	1 day a week is allocated by wild dog controllers to baiting activities An average of 22 baits are laid each week per wild dog controller. Baits are checked and replaced every 14 days for fresh baits and 30 days for shelf stable baits. 80 per cent of baits laid are manufactured.
Bait purchase cost	The average purchase cost per bait (including poison) is $2.50.
Travel cost	3 baiting sites are visited by wild dog controllers on each day allocated to baiting.

Source: MJA analysis

One of the key differences between baiting and trapping is that trapping is more labour intensive (Table ).⁴ Setting traps involves a larger amount of initial preparatory work in establishing the appropriate place to set the trap relative to baits. Additionally, baits placed in one area (e.g. a ‘bait run’) can range from 8 to 40 baits in one day, whereas trapping activities in one area may only involve a few traps. Baiting does require significant administration to seek approvals and undertake large scale neighbour notification processes; warning signs must also be erected prior to baiting and remain for 28 days post baiting on all access points to the baiting sites.