Economic Impacts of Salmonella Dublin in Dairy Farms: Panel Evidence From Denmark

Abstract

Zoonotic1 livestock diseases, such as Salmonella Dublin (SDB), have become a major public health concern in recent decades due to their potential to affect animal protein production, trade, human health, livelihoods, and food security (Hennessy and Marsh 2021). The increased globalization, expansion of human populations, intensification of animal production systems, and changes in land use and climate increase the risk of zoonotic diseases spreading as epidemics and pandemics (Leal et al. 2022). This calls for the need for effective surveillance and monitoring systems to mitigate their impact as highlighted during the COVID-19 pandemic. SDB, a strain commonly found in cattle, is one of the most important zoonotic diseases, with rising incidence and widespread antibiotic resistance, which makes it difficult to treat and deadly, killing up to 12% of human infections (Helms et al. 2003; Harvey et al. 2017; Srednik et al. 2021; Velasquez-Munoz et al. 2024; do Amarante et al. 2025). It can remain latent in herds for extended periods in healthy-appearing animals, spreading through infected animal trade, animal contact, or manure, which complicates control and eradication efforts (Nielsen et al. 2004; Velasquez-Munoz et al. 2024). Despite the high prevalence of SDB in cattle (e.g., 60% in Great Britain (APHA 2022) and 18% in the US (Frye 2021)), its role as the main source of human Dublin infections (Helms et al. 2003; Harvey et al. 2017, Velasquez-Munoz et al. 2024) and its capacity to cause severe illness and death in cattle, particularly in calves (Peters 1985; Holschbach and Peek 2018; SEGES 2022), regulatory bodies worldwide have yet to implement adequate measures to control or eradicate SDB in cattle farming.

The absence of effective monitoring systems and the growing issue of multidrug antibiotic resistance in SDB infections present significant challenges for controlling the disease (Harvey et al. 2017). These factors, coupled with the lack of comprehensive farm-level economic estimates, undermine efforts to convince policymakers and farmers to implement stronger regulations and interventions. For instance, while society may aim to reduce human Dublin infections originating from cattle, as seen with initiatives like Denmark's SDB eradication plan in 2008, empirical evidence does not indicate significant farmer participation in SDB reduction efforts, contrary to predictions based solely on cow-level milk yield outcomes (Nielsen et al. 2012a, 2012b; Nielsen et al. 2013), which often fail to account for the complexities of farm management and production behavior (Seegers et al. 2003). Understanding the economic impacts of SDB at the farm level is crucial for identifying farmers' incentives to engage in eradication efforts and for designing effective regulatory policies aimed at mitigating SDB in cattle farming (Velasquez-Munoz et al. 2024). This paper aims to provide the first comprehensive empirical estimates of the economic burden of SDB infections on dairy farms, leveraging a unique panel dataset of SDB antibody test results from all dairy farms in Denmark.

Two strands of literature are relevant for the present study. First, unlike much of the existing literature that primarily focuses on the public health costs of salmonellosis, our study contributes to the agricultural economics literature by specifically investigating the economic impact on farm operations, which has received relatively little attention. Other studies have largely focused on the effects of surveillance schemes not estimating the economic burden and did not focus on SDB specifically (e.g., Ollinger and Bovay 2018, 2020; Ollinger and Houser 2020). Moreover, existing studies on animal farming have been limited by small sample sizes, lack of data on farmers' production and managerial behavior, failure to examine standard measures of economic performance, and reliance on cost-benefit analysis, computer-based simulation models, and optimization methods (Bergevoet et al. 2009; Nielsen et al. 2012a; Nielsen et al. 2013; Ågren et al. 2015; Rasmussen et al. 2024). The study aims to address these limitations by using unique long farm-level panel data covering all Danish dairy herds that contain detailed information on production and economic statistics and registered SDB antibodies in tank milk (Optical Density Counts [ODC]) to approximate SDB infections in dairy cows (Cummings et al. 2018). SDB transmits through feces and the infection with SDB bacteria can cause Salmonellosis with clinical signs of the disease or the cow can be infected without being clinically ill (Nielsen et al. 2004; Holschbach and Peek 2018). If a cow has been infected by the bacteria, it develops antibodies detectable in milk samples when the concentration is high enough. The duration from the first infection until a high ODC can be found in the tank milk samples depends on the epidemiological development in the herd (Nielsen et al. 2004). The eradication of the bacteria depends on biosecurity measures preventing newborn calves and heifers from being infected (Nielsen et al. 2012b).

Specifically, the paper's key contributions are: (i) it utilizes a unique panel dataset of tests for SDB antibodies for the entire population of Danish dairy farms; (ii) it employs rigorous high-dimensional fixed effects regression models, including farm, quarter, and farm by year fixed effects for quarterly analysis, and farm and year fixed effects for annual analysis, that help address all time-invariant and most time-varying unobservable confounding variables, along with other relevant covariates that account for observable confounders; (iii) it analyzes milk yield at the farm/herd level over a long period; (iv) it takes into account farmers' management and production behavior; (v) in addition to milk yield, the most common outcome in previous studies (e.g., Nielsen et al. 2012a); it examines additional range of outcomes, including calf mortality and production costs; (vi) it utilizes more recent data from 2011 to 2021, enabling the analysis of changes in adaptation to and costs of SDB infections since earlier studies; and (vii) in contrast to previous studies, we also conduct a simulation study to estimate the sectoral economic burden of SDB using the lower and upper bounds of the econometric estimates, thereby providing a confidence interval for the reliability of our burden estimation. These factors make the results directly relevant to policy decisions (Seegers et al. 2003).

Second, this study enhances the existing literature on the economic impacts of livestock diseases by incorporating asymptomatic chronic infections into the estimation of disease economic burdens, an area that has been relatively underexplored. Previous research has predominantly concentrated on one-time disease outbreaks (Caskie et al. 1999; Bennett 2003; Pendell et al. 2007; Park et al. 2008; Saghaian et al. 2008; Ihle et al. 2012; Knight-Jones &Rushton 2013; Cairns et al. 2017) and has primarily examined epizootic2 and enzootic3 diseases within the beef (Pendell et al. 2007; Park et al. 2008; Ihle et al. 2012; Knight-Jones and Rushton 2013; Cairns et al. 2017) and poultry sectors (Saghaian et al. 2008; Antunes et al. 2016). These studies have primarily analyzed data on symptomatic individuals or herds, where the visible impacts of disease on health and production are more easily measurable. In contrast, this study emphasizes the significant yet often-overlooked effects of asymptomatic life-long infections on production behavior and decision-making in the dairy sector by utilizing data on both symptomatic and asymptomatic infections (Harvey et al. 2017; Cummings et al. 2018). This dimension is particularly important for SDB due to its One Health implications, as its high prevalence in cattle and antibiotic resistance pose a public health threat, potentially leading to outbreaks and pandemics (Harvey et al. 2017; Velasquez-Munoz et al. 2024).

Incorporating asymptomatic infections into our data improves our understanding of the economic burden of zoonotic diseases. First, despite lacking visible symptoms, asymptomatic carriers can act as reservoirs for pathogens, facilitating disease spread and potential outbreaks (Nielsen et al. 2004; Harvey et al. 2017; Cummings et al. 2018). Second, these infections can significantly affect management practices and economic outcomes; farmers unaware of them may neglect crucial biosecurity measures, leading to unexpected productivity losses, increased veterinary costs, and potentially significant human health impact (Helms et al. 2003; Harvey et al. 2017). We argue that recognizing the prevalence of asymptomatic cases and their potential economic impacts could shift farmers’ perceptions of overall herd health, leading to more effective management strategies and ultimately reducing the economic risks associated with zoonotic diseases. This study specifically examines the impact of SDB infections on farmers' production behavior in the dairy sector, utilizing a unique panel dataset that includes test results for SDB antibodies—including asymptomatic infections—and production statistics from all dairy farms in Denmark.

The remainder of the paper is organized as follows: Section 2 describes the data; Section 3 outlines the empirical strategy; Section 4 presents the main results, including robustness tests; Section 5 details a simulation exercise using Danish data; Section 6 discusses the results; Section 7 presents the policy implications; and the final section concludes the paper.

Several approaches can be used to address the complexity of the relationship between SDB and economic outcomes at the farm level. Previous studies have shown that large parts of the variations in data are farm-specific and therefore should not be attributed to changes in ODC levels but to farm specific characteristics (Nielsen et al. 2012a, 2013). Therefore, we include farm fixed effects as they control for any time-invariant factors leading to unobserved heterogeneity across individual farms. Examples of farm-fixed effects could include the size of the house of operation, location, breed of cattle, management practices, and so forth.

This study hypothesizes that SDB infections in dairy herds adversely affect economic performance across several dimensions. First, we expect that SDB will reduce milk yield, as infections likely impair the health and productivity of dairy cows. Second, we hypothesize that “newly” infected herds will experience more significant reductions in milk yield and calf mortality compared to herds with no infections. Third, we anticipate that SDB infections will increase calf mortality, as compromised cow health and SDB infections in newborn calves are expected to reduce calf survival. Finally, we expect that SDB will raise variable production costs, particularly those related to veterinary and medical services, and biosecurity measures.

All model specifications include farm fixed effect, ${\alpha }_i$, which accounts for time-invariant unobserved differences between farms due to, for example, management differences. The quarter fixed effects ${\eta }_q$ accounts for any secular trends that affect all farms similarly in each quarter (e.g., environmental regulations, credit constraints, or technological progress similar for all cattle farms). To capture potential non-linearities in the fixed effects, almost all specifications include the interaction term “year-of-sample by farm fixed effects”, ${\omega }_{zt}$, that is, allowing for interactions between farm and year fixed effects. The interaction captures any year varying farm-specific systematic change, for example, yearly change in size of a farm, yearly disease outbreak in each farm, and so forth. The natural choice for the interaction effect in a model with farm and quarter-fixed effects would be to interact farm-fixed effects with quarter-fixed effects. However, due to insufficient degrees of freedom, we opted to use farm-fixed effects interacted with year-fixed effects instead. In addition, as robustness checks, we include results using farm-fixed effects interacted with quarter of the year to capture seasonal farm level unobservable as component yield may not just be a function of breed but also changes with weather and feed and throughout the cow's biological cycle. Finally, the error term ${\epsilon }_{iq}$ accounts for unobserved random variations in farms’ economic or production performance across quarters.

As described in the data section above, the quarterly herd-level panel data is used to estimate various specifications of Equation (1). To estimate different specifications of Equation (2), observations are aggregated annually and at the farm level. Both regression equations are estimated using STATA 17.0 software.

The following section presents the major findings from our study on the economic consequences of SDB infection on dairy farms. We present its impact on milk production, calf mortality rates, and farm production costs.

The significant relation between SDB in bulk tank milk and milk yield found in Table 5 can be used to assess the overall income foregone and the results from Table 9 can be used to assess the costs due to SDB. Given the development in SDB prevalence, the most recent of the 10 years is used to assess the industry income foregone and costs presented in Table 9. In total, there were 569,000 dairy cows in Denmark in 2020 with the majority having an ODC equal to zero. The total income foregone is EUR 7.5 million and the total cost for the industry is EUR 5.1 million. The average price for milk is found to be 36.1 Eurocent per kg of milk in 2020, with 90 percent being conventional milk and the remaining 10 percent being organic milk.

The industry costs found in Table 9 can be assessed to specify the contribution from different cost components. The contribution from the cost components presented in Figure 3 shows that the increase in feed costs is making up the largest share of the increase in costs related to SDB. The 95 percent confidence interval is presented with the vertical lines in the bars and represents an indication of statistically significant parameter estimate at 5 percent level, but in the industry costs in Table 9, the costs are included if the correlation is significant at the 10 percent level.

If the industry costs are assessed by the non-linear (binary) model based on the same number of cows in 2020 the total income foregone is 6.2 million EUR, that is, the estimated income foregone would be slightly lower in the non-linear model but still in the same range, whereas the total costs would be 6.1 million EUR per year and hence higher than in the model with discrete ODC levels (Table 9).

From a regulatory perspective, understanding the economic impact of SDB on dairy production is key to determining appropriate measures. If the disease causes significant losses, information on its burden and eradication strategies could be effective. If losses are minimal, stronger incentives may be needed to address its broader One Health implications. This paper presents the first comprehensive empirical estimates of SDB's economic impact on dairy farms using a unique panel dataset of Salmonella antibodies from milk deliveries across all Danish dairy farms.

The study finds that milk yield, calf mortality, and total variable costs are associated with SDB at the herd level. We estimate that herds infected with SDB may experience a reduction in milk yield of up to 13 kg per cow per month, which represents a 1.6% loss based on an average milk yield of 833 kg per cow per month. This decrease in milk yield translates into a financial loss of EUR 5 per cow per month, or EUR 58 per cow annually. For a herd of 200 milking cows, this could amount to a monthly loss of EUR 970. Moreover, if a herd experiences 25 < ODC < 40 for a period of 6 months, it is estimated to reduce income by EUR 5600. These findings underscore the importance of preventing and controlling SDB in dairy herds to ensure farmers can maintain stable incomes. However, the financial impact might not be immediately noticeable to farmers due to the variability of production factors, as well as differences in how the effects manifest between individual farms. Although the magnitude is less significant compared to earlier studies (e.g., Peters 1985; Velasquez-Munoz et al. 2024), the study also finds a strong correlation between higher ODC levels and increased calf mortality. This may provide one explanation for why farmers often fail to act when ODC levels are low. Since calf mortality is the most visible and immediate concern for farmers, the absence of this effect at lower ODC levels could delay intervention, even though reduced milk yield is already affecting farm profitability.

The study also estimates that the increased variable costs associated with SDB infections can be significant for farmers. Our simulation results indicate that the lower and upper bounds for economic burden estimates expand at higher ODC counts. For a typical farm with 200 milking cows, a small level of ODC (between 0 and 10) is estimated to result in increased variable costs of EUR 6700 per year, compared to a herd without SDB. A larger level of ODC (between 25 and 40) is estimated to be associated with an additional variable costs of EUR 11,300 per year, compared to a herd without SDB. These increased variable costs correspond to 1%–3% of the total variable costs per cow per year, which is estimated to be EUR 2800. This highlights the importance of taking measures to prevent and control SDB infections to minimize the additional variable costs for farmers. However, it is important to note that these estimates should be interpreted with caution, as the causal relationship between SDB infections and production outcomes has yet to be precisely determined.

Direct comparisons with other studies are challenging, but the closest study by Nielsen et al. (2012a) differs from our findings for several reasons. (i) Our study uses data from 2011 to 2020, while Nielsen et al. (2012a) analyzed data from 2005 to 2009. (ii) Nielsen et al. (2012a) analyzed cow-level milk yield data, while our study emphasizes herd-level data encompassing a broader range of economic outcomes. This approach enables us to explore the complexities of farm management and production behavior (Seegers et al. 2003). (iii) Their study examined a small sample of Holstein herds with a sharp rise in ODC levels, compared to a control group with stable ODC levels, which may explain some of the differences. Although we also analyze “newly” infected herds, we find no significant differences in milk yield compared to other (potentially non) infected herds. Another possible explanation is that farmers may have adopted more mitigation measures over time, resulting in lower production losses but higher costs, as reflected in our findings. Unlike Nielsen et al. (2012a), we also estimate link between SDB infections with calf mortality and production costs. On the other hand, Nielsen et al. (2013) predicted higher production costs from SDB infection using a simulation model. However, unlike our study, they relied on simulations rather than statistical analysis of empirical data, which may reduce the reliability of their estimates.

Compared to other animal diseases, SDB is associated with statistically and economically significant reductions in milk production even at low infection levels, where cows act as carriers without necessarily exhibiting acute symptoms. Other diseases, such as paratuberculosis (caused by Mycobacterium avium), require higher infection rates at advanced stages to produce similar economic impacts (Garcia and Shalloo 2015). Mastitis primarily causes milk yield reductions at later stages or during severe outbreaks, with economic losses escalating with the severity of the infection (Seegers et al. 2003; Hadrich et al. 2018), and none of these are zoonotic. Diseases like Bovine Leukemia Virus and Foot-and-Mouth Disease also result in significant losses, but typically at more advanced stages of infection (Knight-Jones &Rushton 2013; Norby et al. 2016). Interestingly, Bovine Leukemia Virus may even unexpectedly increase milk yield (Yang et al. 2022). However, the early impact of SDB underscores the critical need for early detection and control strategies to minimize both production losses and its significant public health threat, as the disease kills 12% of human infections (Helms et al. 2003).

Even with access to a considerable amount of high-quality data, capturing the relationship between SDB infection in cows and production losses is challenging for several reasons. First, we use SDB antibodies in tank milk (a herd-level observation) as an indicator of SDB infections (a cow-level observation), which makes establishing the time lag between SDB introduction into a herd and the rise in ODC in tank milk complex. Although ODC is a useful proxy for infection rates, it is not a perfect measure. For example, bulk milk testing rarely produces false positives, though cross-reactivity or contamination can lead to misleading results. On the other hand, dilution from the herd's milk and weak antibody responses in chronic carriers may result in false negatives (Cummings et al. 2018). Therefore, ODC should ideally be used alongside other diagnostic and biosecurity measures for a more complete understanding of herd health. Moreover, SDB may not infect all animals in the herd simultaneously, meaning that the ODC rise in tank milk could occur before all cows are infected. There may also be a delay between cows being infected with SDB and an increase in ODC levels, as initial infections may involve only a few cows, making herd-level declines in milk yield less noticeable.

We tested several lag and lead models, but none resulted in better model fits than the model without these variables. Furthermore, because we lacked daily milk yield records, we relied on monthly deliveries to the dairy. Our initial monthly data analysis found no evidence of a lead or lag relationship between changes in ODC and milk yield, consistent with Nielsen et al. (2012a). However, the relationship between ODC and production output could be uncertain and may not be fully captured by our herd-level models. We also explored potential links between “new” SDB infections and production outcomes, but defining new infections proved challenging and beyond the scope of this study. Thus, the production loss figures are rough estimates of the impact of actual infection levels in herds.

Second, the study could analyze the ODC using either monthly or quarterly data, each with its advantages and disadvantages. Monthly data offers greater precision in time correlations but may introduce uncertainty in data collection, such as variations in milk delivery frequency. Since ODC data are only available quarterly, the study uses the same frequency to maintain compatibility between the two variables and reduce uncertainty around milk delivery frequency. However, this approach loses some information on month-to-month variation due to the aggregation into quarterly data.

Third, using the fixed effects model provides strong evidence of the explanatory power of ODC on economic outcomes. However, it has limitations, such as not identifying which farm-specific factors, like farmhouse age, farm size, breed, or the use of automated milking machines, most influence milk yield loss, as these are included in the farm fixed effect term. Since this analysis focuses on the relationship between ODC and farm economic outcomes, future studies could explore other approaches to investigate the impact of other factors on farm economic outcomes, in addition to ODC.

This study provides new evidence on the effects of SDB on farm economics, but future research is needed to establish causality, enhance the robustness of the findings, and offer more detailed insights at the cow level for a deeper understanding of the effects.

The study underscores the economic challenges posed by SDB infections in dairy farming, affecting milk yield, mortality rates, and operating costs. Although impacts vary across herds and may not be immediately noticeable, there is an urgent need for early management of SDB, even at low prevalence levels. The asymptomatic nature of SDB often leads farmers to underestimate its economic consequences, diminishing their incentives to comply with national eradication efforts. Farmers may overlook low infection levels for other management priorities, leading to higher costs in the long run.

To tackle these challenges, we propose the following measures: (i) Farms experiencing significant losses from SDB should receive targeted information on the economic impacts and effective eradication strategies. For those with minimal losses, considering that some farmers may not perceive the economic burden of SDB, policy initiatives should offer substantial incentives, such as subsidies for prevention efforts and early herd testing, or adjusted milk pricing for farms with prolonged SDB infections. (ii) Existing arbitrary ODC threshold-based regulations should be reassessed, as even low levels of SDB can are associated with significant loss of milk yield. Last, implementing a comprehensive monitoring system that includes regular testing is crucial for effective SDB eradication. This system should gather both herd-level and individual cow data to ensure a complete understanding of disease dynamics and accurate estimation of the sectoral burden. The policy implications can be wrapped up by stating that, although significant correlations are found at the herd/farm level between ODC and milk yield, the magnitude of the associations with calf mortality and variable costs is too low to effectively incentivize farmers to prevent and eradicate SDB infections and curb human infections.

This study uses a long panel dataset on production statistics and levels of SDB antibodies in tank milk (represented by ODC levels). By applying high dimensional fixed effects models, the study minimizes endogeneity concerns and finds a statistically significant correlation between milk yield and SDB infections. Specifically, herds with non-zero ODC levels in tank milk experience a monthly milk yield loss of approximately 8 kg per cow compared to herds with ODC of 0. For herds with ODC levels above 55, the estimated milk yield loss increases to 13 kg per cow per month, which corresponds to an income loss of EUR 58 per cow per year, assuming a milk price of 40 Eurocents per kg. These losses, while smaller than those found in previous studies by Nielsen et al. (2012a, 2013), highlight that even small increases in ODC are associated with productivity loss.

Interestingly, the study did not find a significant association between ODC and milk yield for “newly” infected herds, contrary to Nielsen et al. (2012a) findings. This suggests that SDB is associated with milk yield loss as soon as ODC levels are positive, such as below 10 ODC, not only at higher thresholds, which indicates the need for more stringent rules than existing arbitrary threshold-based regulations. This finding has policy implications because the current Danish surveillance scheme is implicitly built on the notion that “newly” infected herds are going to be incentivized by the economic consequences of having an ODC above 25, whereas we find a correlation for ODC > 0.

In addition to milk yield losses, the study also documents small increases in calf mortality, particularly among bull calves, correlated with ODC levels above 0. For example, herds with ODC above 55 experienced 14 extra dead heifer calves per 1000 live-born calves compared to herds with ODC of 0. Although this increase is small, the correlation with higher ODC levels, including “new infections”, may explain farmers’ delays in taking action when ODC levels are low. Since calf mortality is the most visible concern, its absence at lower ODC levels may delay intervention, even though reduced milk yield is already affecting profitability.

Moreover, the results suggest that “new” SDB infections are associated with increased calf mortality, but not with milk yield, indicating that calves may be more vulnerable to “new” SDB infections than cows. In addition to herd-wide preventive measures, targeted interventions such as vaccination are recommended to reduce calf mortality (Cummings et al. 2019).

It is important to note that this study does not isolate the effect of SDB from other changes that are implemented at the same time at herd level. For instance, if farmers are successful in mitigating the milk yield loss due to changes in hygiene level or increased overall management performance, then the isolated effect of SDB in a herd could be larger than the estimates provided.

Ultimately, although the herd/farm level correlations between ODC and milk yield, calf mortality and variable costs are attributed to SDB infections, the magnitudes of the economic losses are too small to be easily detected by farmers or to sufficiently incentivize them to effectively prevent and eradicate SDB infections and curb human infections. Stronger incentives need to be implemented to drive effective actions.

DB and JVO conceptualized the study. JVO acquired the data. DB developed the empirical strategy, curated and visualized the data, and conducted the formal econometric analysis, while JVO performed the simulation study. DB drafted the manuscript, while JVO contributed through editing, validation, and critical review. Both authors discussed and validated the empirical results and approved the final version of the manuscript.

The authors declare they have no competing interests.