Tag: nudge

Recipient Type and the Effectiveness of Informational Campaigns: The Case of Meat

A picture of meat on the table table, indoor, sitting representing Informational campaigns that affect meat consumption

While global population growth has been accelerating during the last decades, the number of humans currently living on the planet is dwarfed by the amount of farm animals alive at any time, and even more by the quantity we slaughter for meat every year. According to the latest FAO statistics, this latter number is estimated at around 75 billion. Even ignoring animal welfare, this is severely affecting the health of the planet and our own. What should be done about this?

Externalities of Meat Consumption

Mankind has been butchering and eating animals for at least 3,4 million years (McPherron et al., 2010). Evolutionary biology theories claim that complementing our diet with meat contributed to the spectacular growth of our brain (Fonseca-Azevedo et al., 2012). Anthropological theories suggest that the necessity of hunting drove the development of tool building, language and social structures. The domestication of animals (and plants) around 10,000 years ago led to a jump in the history of civilization. In other words, eating meat is a large part of what made us human. However, during the last century, we took this to unsustainable levels. All in all, the agricultural sector accounts for 25 to 30% of global CO2 emissions, second only to the energy and transport sector, and 60% of non-CO2 emissions, in particular methane, which is much more efficient than CO2 at warming up the planet. A third to half of these emissions, depending on whether or not we include the share related to land use, comes from livestock production. Large scale factory farms, which cater to the ever-increasing global demand for cheap meat, are also responsible for other externalities, including distorted resource use (in particular of water and fertile land); local pollution of air and waterways, with consequences for neighbouring ecosystems and human health; abuse of antibiotics, which threatens their effectiveness with dramatic implications for the whole spectrum of modern medicine. The cheap and overabundant animal products with worsened nutritional properties, which result from these production methods are also behind the epidemic of “welfare diseases” such as diabetes, cardiovascular conditions and some types of cancer (Mozaffarian, 2016).

So, what should we do about this? Economics is very clear on this point. In the presence of externalities, market prices do not reflect social costs; therefore, the market mechanism fails, and decisions taken on the basis of these prices are suboptimal. If applied to meat consumption, this principle implies that, first of all, consumers and producers must pay for the emissions (and other externalities) they cause. Today’s carbon pricing systems, whether in the form of a tax like in Sweden or tradable emission permits like in EU, exempt the agricultural sector for various reasons. Moreover, as already mentioned there is more to meat than carbon emissions. Another FREE brief (Perrotta, 2011) makes the case for a meat (consumption) tax. Multiple teams of researchers (Wirsenius, Hedenus, and Mohlin, 2011; Edjabou and Smed, 2013; Gren, 2015; Andersson, 2019) have come as far as to compute the optimal level of such a tax, in different contexts and under different assumptions. There are also drawbacks to this approach, though. Climate-change curbing policy is in general an area where policy makers at all levels find it hard to converge to policies of strong incentives, such as taxes and regulation. Interventions targeting food production or dietary choices, in particular, are likely to face strong opposition from producers and consumers alike. It is therefore worth considering the alternative – or at least complementary – strategy of information and awareness campaigns.

The Power of Information

Given that a climate policy agenda of strong incentives is so fraught with obstacles, the potential for information to spark voluntary action would be very valuable. There is a catch here, however. Information about the benefits of an action often fails to encourage that action. Consider the case in point: for decades now, we have observed a persistence and increase of meat eating despite mounting evidence and widespread information on the ills of meat production and consumption. Indeed, this well-known weakness of informational interventions has contributed to the rising importance and application of alternative approaches. One example is the popularity of the so-called nudges (Thaler and Sunstein, 2009), modifications in the choice architecture that can subtly push agents towards an action without actually limiting the available alternatives. There is ample research on where and why the chain from information to action might get interrupted, and established evidence that the effectiveness of information depends on a variety of factors such as recipients’ prior beliefs, the sender’s credibility, and the non-informative content of the message, such as the emotional evocativeness of imagery (see a survey in DellaVigna and Gentzkow, 2010). Taking a step back to the stage before, namely the question whether information does reach the intended beneficiaries in the first place, at least three aspects of this have been investigated: limited attention, active avoidance, and selective retaining of information on the part of the recipients. In a new working paper (Berlin and Mandl, 2020), we investigate the role of individual type for selective information retention. We ask whether certain types of agents, in our case vegetarians, retain more of the information they are exposed to, even when exposed to a similar context and the same incentives to retain information as everyone else (so that hopefully the competing channels of limited attention and active avoidance can be neutralized). This has relevance for the possibility of tailoring the policy message, similar to the marketing theories of market segmentation. In contrast to well-developed marketing practices in the private sector, this potential has so far not been exploited in policy design. To the best of our knowledge, this mechanism has not been investigated in a real-life incentivized setting outside the lab before.

Natural Experiment in Class

We exploit a natural experiment in the context of higher education. A class of college students was assigned an essay about their plan for a Christmas dinner menu, after being exposed to a lecture and reading materials on the externalities of meat production, so that they could decide to make use of this information. The essays were to be written in randomly assigned groups of three, making the type combination, i.e. the presence of one or more vegetarian group members, a random group characteristic. We hypothesize that there is a difference in how carnivores and vegetarians deal with the provided information about the food industry. In particular, we test whether groups that include a vegetarian student recall a larger share of the information than groups made up only of carnivores. The essay was mandatory, and moreover it awarded study credits toward the final grade of the course (10/100 points). This constitutes a sizeable incentive and possibly provides a stronger motivation for information retention as compared to the average monetary rewards which lab experiments rely on. To measure the share of information retained, we preregistered a list of 30 words in both English and Swedish related to the learning outcomes of the lecture. We then used a script to measure how many of the 30 words appear in each essay. We call this number the essay’s score, separate and independent from the teacher’s assigned grade, which is of relevance for the student. The teacher-assigned grade, reflecting general comprehension of the topic rather than just the presence of keywords, is expected to be correlated with the score, but not perfectly. We also expect the grade to capture the ability of the students to a higher degree compared to the score, as the automatized word count fails to consider the context in which the words are mentioned.


Figure 1. Group score by treatment status

Source: Berlin and Mandl (2020).

On average, groups including a vegetarian student scored higher (4.8) than groups with all meat-eaters (4.3), but not significantly so. The estimated Cohen’s d (0.347), a standard measure of effect size used to indicate the standardized difference between two means, is much smaller than the minimum detectable effect in our sample, which we estimated at 0.8. In other words, we do not have the statistical power to either accept or reject the null hypothesis. The reason is that the treated group displays larger variation in score outcomes, possibly due to the smaller than anticipated sample size: only 11 students out of almost 300 identified themselves as vegetarians or vegan (non meat-eaters), which is a much smaller proportion than what the latest survey of young adults in Sweden estimates (17%, Djurens Rätt, 2018).

Looking beyond the mean at the details of our data reveals an interesting pattern. As the Figure shows, the distribution of achieved scores among the vegetarian groups is bimodal: a lower-level concentration of scores is close to the mode of the control distribution, but there is an almost as large mass at a higher level. This might suggest that, quite understandably, (attention and) performance, in terms of recall, is affected by several factors beyond the type. In other words, not all the individuals with the relevant type display increased retention of information. While many vegetarians remain close to the mode for the meat-eating type, a large fraction obtains double the score, suggesting a substantial though heterogenous increase in the retention of information.

We also use regression analysis in order to control for potential omitted variables and net out some of the variation in the score data that is not related to our variable of interest (such as group size and ability). Robustness checks were performed with different specifications and alternative outcome variables, but the main conclusion remains the same: mean performance, in terms of information retention, is higher for the vegetarian type but not significantly so. However, these results should not be interpreted as a rejection of our original hypothesis about the importance of type for information retention, as our analysis is empirically underpowered due to the low number of vegetarians in the sample. More importantly, the method we propose is highly appropriate, easily replicable and cheap.


Information interventions are low-cost and can be effective. Understanding how they can be tweaked for best effect is an area of crucial research interest, in particular for such an area as climate-change curbing policy. We provide an easy and cheap method to investigate this further and hope that more future research will pursue this avenue.


  • Andersson, Julius J., 2019. “The ‘meatigation’ of Climate Change: Environmental and Distributional Effects of a Greenhouse Gas Tax on Animal Food Products.” London School of Economics
  • Berlin, Maria P. and  Benjamin Mandl , 2020. “Selective attention and the importance of types for information campaigns”, SITE Working Paper Series. 53.
  • DellaVigna, Stefano and Matthew Gentzkow, 2010. “Persuasion: empirical evidence.”, Annu. Rev. Econ. 2 (1), 643–669.
  • Djurens Rätt, 2018. “Opinionundersökning, Våren 2018.” Novus.
  • Fonseca-Azevedo, Karina and Suzana Herculano-Houzel, 2012. “Tradeoff between brain and body mass.” Proceedings of the National Academy of Sciences. 109 (45), 18571-18576
  • McPherron, Shannon P. et al., 2010. “Evidence for stone-tool-assisted consumption of animal tissues before 3.39 million years ago at Dikika, Ethiopia.” Nature 466857–860.
  • Mozaffarian, Dariush, 2016. “Dietary and policy priorities for cardiovascular disease, diabetes, and obesity: a comprehensive review.” Circulation. 133(2), 187–225.
  • Perrotta, Maria, 2011. “Tax Meat to Save the Baltic Sea.” FREE Policy Brief Series.
  • Säll, Sarah and Ing-Marie Gren, 2015. “Effects of an environmental tax on meat and dairy consumption in Sweden.” Food Policy. 55, 41-53.
  • Thaler, Richard H. and Cass R. Sunstein, 2009. “Nudge: Improving Decisions About Health, Wealth, and Happiness.” Penguin Group.

Disclaimer: Opinions expressed in policy briefs and other publications are those of the authors; they do not necessarily reflect those of the FREE Network and its research institutes.

The Georgian Tax Lottery of 2012 – A Quantitative and Qualitative Evaluation

20191104 The Georgian Tax Lottery of 2012 FREE NETWORK Policy Brief Image 01

This policy brief is based on preliminary findings of research that assesses the 2012 Georgian Tax Lottery by Larsen et al. (2019). Tax lotteries are seen as a way to relatively easily augment public revenue while also increasing compliance. Tax lotteries are constructed so that consumers are nudged to ask for a receipt when making a purchase. This receipt contains information which can also be used as a lottery ticket with the possibility of winning prizes. Such tickets also leave traces of transaction records that allow revenue authorities to audit vendors. Given this background, the aim of this paper is to provide a broad, multi-methodological and socio-economic assessment of Georgia’s tax lottery experience in 2012.


A well-designed tax system improves economic efficiency, facilitates economic growth and social welfare, (Besley & Persson, 2013). Yet, curbing tax evasion remains one of the key challenges for policy makers, and institutions in charge of revenue administration are experimenting with diverse set of instruments to increase tax compliance and thus revenue.

In addition to the traditional audit-sanctioning mechanism, the taxation literature emphasizes the role of consumers in facilitating tax compliance of businesses. The government can create direct monetary incentives for consumers to request receipts. Turning a receipt into a lottery ticket with a chance of winning a pre-determined prize is an example of such an incentive. The tax lottery motivates and rewards those consumers who become part in the efforts to fight tax evasion by requesting receipts while making purchases. Given that audit-sanctioning mechanisms are very costly for the government, clever usage of a “zero cost policy”, such as tax lotteries, might be advisable (Fabbri & Hemels, 2013).

The aim of this paper is to provide an assessment of the Georgian tax lottery experience in 2012 using both quantitative and qualitative methodologies. The two methodological approaches complement each other and help to investigate the tax lottery from different angles.

The Georgian Tax Lottery

The Georgian Revenue Service (GRS) introduced a tax lottery starting in spring 2012, which was planned to run until January 1, 2013. The aim of the lottery was to popularize the already introduced General Packet Radio Services (GPRS) -based cash registers and make sure that they were used by vendors. Such registers would allow the GRS to gather information about business activities online daily. This, in turn, was due to an effort to fight the shadow economy and be able to audit business revenue, when payments were made by cash. The lottery would thus motivate consumers to ask for receipts. As a communicative resource, the lottery aimed to increase awareness of asking for receipts, as well as to develop a positive attitude in Georgian society towards GRS in the background of harsh fiscal reforms.

In order to participate, customers had to buy goods or services from a vendor who had a GPRS-based cash register. The receipt could be checked for win immediately by mobile phone. The Georgian Tax Lottery was a chance to win money for every customer purchasing anything from groceries, to shoes and hair care. The winning prizes were 10, 20, 50, 100, 10,000 and 50,000 GEL[1]. The 10,000 GEL prizes were awarded once a month while 50,000 GEL prizes were given quarterly.

The lottery ended prematurely on grounds of inefficiency on November 12, 2012 when a new government was elected.

Multi-Method Approach

For the assessment of the tax lottery in Georgia, we employed a multi-method approach combining a qualitative assessment built on an ethnographic approach with quantitative regression-based methods; following the ethnographic approach, we collected opinions, experiences, and views on the tax lottery from the perspective of participating and non-participating businesses, consumers as well as other stakeholders.

The quantitative assessment of the paper investigates whether the existence of the lottery affected businesses’ total revealed turnovers through the facilitation of a receipt-requesting norm. The data for the quantitative analysis conducted in this paper was provided by the GRS. The latter was collected from the daily reports of the GRS system, for two years, 2012 and 2013. The data includes variables, such as the unique cash register identifier, the year and the week of a purchase and address (city and municipality) and the total turnover of the cash register reported through GPRS. GRS also provided the dataset with detailed information on winning tickets. The latter includes daily information on the number of winning tickets and the aggregate daily monetary amount of the prizes.

Three different specifications of linear regression models were run separately on the aggregate country level data. The model-specifications differ in a way that each uses different dependent variables – aggregate weekly sales, average weekly sales per register and number of registers reporting any sales.

Preliminary Results

Table 1: Regression Results of the aggregated analysis on a country level

As may be inferred from the country level regression results reported in Table 1, for all the econometric specifications the ‘lottery’ variable is significant at 1% level. The regression results show that during the weeks of the lottery (weeks 16-46) the aggregate weekly sales are on average 33,363 GEL higher than in the non-lottery weeks (11% more than in non-lottery weeks, based on the log linear model). When looking at the year effect of 2012 in non-lottery weeks, the effects are positive, significant, and, on average, amount to 38,813 GEL. This means that aggregate weekly sales in the non-lottery weeks of 2012, exceed aggregate weekly sales in 2013, on average, by 38,813 GEL. While in this simple model we do not explicitly control for the macroeconomic environment, GDP in 2013 grew by 3.4% while inflation stood close to 0%. These macroeconomic outcomes strengthen predictions of the econometric analysis.

When looking at the average sales per register as the dependent variable instead of aggregate weekly sales, the results are compatible with the results of the first model. There is on average a 282 GEL (7.7%) increase in average turnover during the lottery weeks compared to the non-lottery weeks; and average weekly sales in non-lottery weeks of 2012 exceed average weekly sales in 2013 by 458 GEL, on average. In addition, the positive effect and significance of the year 2012 variable shows that controlling for the non-lottery weeks, something was still driving sales up. This could be the long-term effect of the lottery weeks that continued even after the termination of the lottery; hence some evidence of habit formation.

A similar regression is done with the weekly number of cash registers reporting their income as a dependent variable. The outcome illustrates that during the lottery weeks of 2012, the average number of reported cash registers is 3,199 units (4%) more than those in non-lottery weeks, which is quite compatible with the results reported by the first and second regressions.


Despite seemingly positive results, the lottery was prematurely terminated after parliamentary elections in November 2012. Interviews with stakeholders revealed that the public budget that was allocated for the lottery was deemed insufficient to keep the chances of winning high enough and therefore interest and participation from public had decreased significantly from around 2 mln out of 2.5-2.8 mln receipts checked daily in the first months of the lottery to only 300,000 by the end of the lottery. However, there was a lack of financial resources or interest from the new government to invest additional resources to increase the budget and effectiveness of the lottery.

Regardless of its premature termination lottery itself was thought to have influenced social norms and also started a discussion about tax compliance. The tax lottery also aimed to improve citizens’ attitude towards the GRS. A qualitative analysis, based on multi-ethnographic approach through which we have collected media articles, reports, and other materials expressing views on the Georgian tax lottery, however, showed that strategies of “love and fear” are difficult to make work in combination, and we find it hard to say that citizens’ views of the GRS improved due to the lottery itself. Perhaps even the contrary could be proposed. In terms of an increased trust to the GRS, we conclude with our methodological point that a tax lottery cannot be assessed as an isolated event. Previous and other activities that the revenue services engage in that have an impact on taxpayers and on societal tax, compliance have to be taken into consideration. Fear and unjust treatment especially linger in people’s perceptions.


  • Besley, T., & Persson, T. (2013). Taxation and development. In Handbook of public economics (Vol. 5, pp. 51-110). Elsevier.
  • Fabbri, M., & Hemels, S. (2013). ‘Do you want a receipt?’ Combating VAT and RST evasion with lottery tickets. Intertax41(8), 430-443.
  • Larsen, L., Arakelyan, R., Gogsadze, T., Katsadze, M., Skhirtladze, S., & Muench, N. (2019). The Georgian Tax Lottery of 2012. A Multi-Methodological Assessment. International School of Economics at TSU, Tbilisi, Republic of Georgia.
  • Marcus, G. E. (1995). Ethnography in/of the world system: The emergence of multi-sited ethnography. Annual review of anthropology, 24(1), 95-117.

[1] The exchange rate for a Georgian Lari, GEL, is about 3.0 GEL to 1 EUR.