On Feta and Fetta: Protecting EU geographical indications in Australia

The European Union (EU) protects about 1500 regional specialty foods such as Feta cheese and Gouda Holland. However, this geographical indications (GIs) protection does not automatically extend to other jurisdictions. This article tests a theory of EU demands for GIs in EU– Australia trade negotiations, using newly coded data on EU GI names in Australian supermarkets. Focusing on cheese, it confirms that the EU seeks protection for GIs where the authentic EU GI products are well established. The demand for protection does not seem to be driven primarily by current non-GI uses of GI names, since a qualitative analysis reveals few product names that would be banned if EU demands are granted by Australia. These findings imply that Australian negotiators and producers should be less afraid of losing currently established generic food names— a fear that is especially present in the United States, which has strongly opposed the protection of GIs worldwide.


| I N T RODUC T ION
Geographical indication (GI) protection has been a red line in EU trade negotiations (DG AGRI, 2012;Hogan, 2019;Moir, 2016): the European Union (EU) seeks protection for GIs with high global sales and from the Mediterranean countries where gastronationalism is relevant (Huysmans, 2020;Wanat & Hanke Vela, 2019).As part of EU-Australia trade negotiations, the EU has asked Australia to protect 172 EU GIs.However, while we know that global sales drive demands for protection, we do not know whether sales of EU GIs or identically named non-GI products in the target country drive these demands. 1  This article contributes to the literature on GIs and trade policy by extending an argument by Meloni and Swinnen (2018) on the historical origin of GIs to current-day EU trade policy on GIs.Empirically, it innovates by integrating EU demands for GI protection with Australian supermarket data and EU export data.It explores the important question of whether the EU demands protection mostly for well-established GIs or for GIs that currently face competition from identically named non-GI products.In other words, are the EU's demands mostly proactive and about foreclosing future imitations, or are they focused on clawing back currently generic food names?
The so-called 'war on terroir' (Josling, 2006) means that research into GIs in trade agreements is highly policy-relevant.Whereas the EU considers protection of at least some of its GIs in partner countries a red line for concluding a trade deal, the USA vehemently opposes GI protection (Goldberg, 2001;Hughes, 2006;Informa Economics IEG, 2016;Livingstone, 2017;Mancini et al., 2016;O'Connor & Bosio, 2017;Raustiala & Munzer, 2007;Watson, 2016).The US actively interferes in negotiations between the EU and third countries: in 2020, the US Department of Agriculture supported the Consortium for Common Food Names (CCFN) to file opposition to 12 GIs the EU asked Australia to protect (US IPEC, 2021).Both the CCFN and Dairy Australia have even objected to the registration of new cheese GIs within the EU. 2pposition may be particularly strong from producers of European descent, who feel they should be allowed to use historical names for recipes of ancestral origin.
This article studies EU-Australia negotiations for a free trade agreement (FTA).Like the USA, at the World Trade Organisation (WTO) Australia was against increased multilateral protection of food GIs (Kneller, 2020;Raustiala & Munzer, 2007;Van Caenegem et al., 2014)see Addor and Grazioli (2005) for a discussion on GIs and the WTO agreement on trade related aspects of intellectual property agreement (TRIPS).In August 2020, the Australian pork 1 The remainder of the paper sometimes uses the term 'imitation' instead of 'identically named non-GI products'.The term imitation is used from the perspective of the EU GI holders.Non-EU users of names that are protected as GIs in the EU may be using these names in good faith as generic descriptors of a product type.
industry federation stated that 'GIs are a tool of protectionism that are utilised to reduce competition and limit innovation.GIs have no place in FTAs.[…] Widespread adoption of foreign GIs will hurt Australian farmers and small goods manufacturers and cause confusion amongst Australian consumers' (Australian Pork, 2020: 9).
If Australia agrees to protect EU food GIs, 3 it would be a major win for the EUconsolidating its growing worldwide success illustrated by the inclusion of GIs in the EU-Canada Comprehensive Economic and Trade Agreement (CETA) and in a GI-specific deal with China (Hanke Vela & Momtaz, 2019;Huysmans, 2020;Livingstone, 2017).It would also have implications for international trade in the post-Brexit world.One of the reasons for Brexit was for the UK to have greater leeway in concluding FTAs (Foster & Brunsden, 2020).While it was part of the EU, its trade policy was set in Brussels.The EU's insistence on GI protection was one of the reasons the Transatlantic Trade and Investment Partnership (TTIP) did not materialise (Matthews, 2016;Prescott et al., 2020).During the Brexit transition period, the UK appeared to want to dial back its protection of EU GIs (Rankin, 2020). 4If, after Canada, Australia also agrees to protect EU food GIs, this would reduce US pressure on the UK to relax protection of EU GIs.
EU-Australia negotiations over GIs are also worth studying because Australia might eventually seek to capitalise on EU demands by setting up its own food GI system (Van Caenegem & Nakano, 2020;Zito, 2021). 5This would be consistent with a trend where EU trading partners have set up their own GI systems in response to EU trade agreements (Biénabe & Marie-Vivien, 2017;Covarrubia, 2011;Kimura & Rigolot, 2021;Park, 2020).As part of the FTA, the partner GIs can then immediately be protected on the EU market as well.Past examples include Kobe Beef from Japan, Red Ginseng (Goryeo Hongsam) from Korea, and Phú Quốc fish sauce from Vietnam.Suggesting that Australia might indeed establish a sui generis food GI system, it held a public consultation on a 'Possible New Geographical Indications Right' from 4 September to 30 November 2020.
This article contributes to the growing empirical literature on GIs and trade policy (Agostino & Trivieri, 2014;Curzi & Huysmans, 2021;Curzi & Olper, 2012;Raimondi et al., 2020;Sorgho & Larue, 2014) by integrating Australian supermarket data with EU export data and demands for GI protection.Theoretically, it extends an argument by Meloni and Swinnen (2018) on the historical origin of GIs to current-day EU trade policy on GIs.By collecting data prior to the conclusion of a trade agreement, new light can be shed on the determinants of demands for protection.Focusing on cheese and processed meat GIs, data were collected on the availability of EU GI products and identically named non-GI products in the leading Australian supermarket (Woolworths).To the best of my knowledge, this is the first paper on the determinants of GI protection in FTAs to use supermarket data.The supermarket data was complemented with trade data at the CN8 level.
A qualitative assessment of identically named non-GI products currently being sold in Australia complements the quantitative analysis.The protection sought by the EU is strong: it would ban descriptors such as 'Feta-type cheese made in Australia' or 'Australian cheese in the style of Roquefort'.Translations, for example 'Parma ham' for 'Prosciutto di Parma', would 3 In round 11 of the EU-Australia negotiations (June 2021), Australia reiterated that it 'will not agree to protect EU GI terms unless the overall deal is in our interests'; see https://www.dfat.gov.au/trade/agree ments/ negot iatio ns/aeuft a/news/austr alia-eu-fta-repor t-negot iatin g-round -eleve n-1-11-june-2021 4 There have also been calls from Australian producers for the UK to dial back GI protection.For instance, Wine Australia called for Prosecco to be allowed as a grape varietal name.Rock lobster exporter GFC stated that 'Australia needs to ensure that text regarding rules of origin, geographical indicators (GIs), and vessel ownership conditions do not prevent Australian seafood producers from exporting product under the preferential terms of the FTA'.Public submissions related to UK-Australia FTA negotiations are available at https://www.dfat.gov.au/trade/agree ments/ negot iatio ns/aukft a/submi ssions also be protected.Surprisingly, the qualitative analysis reveals that currently very few 6 names are being used in Australian supermarkets that would likely no longer be allowed under the EU's demands.One example is Kransky sausages, a translation of the Slovenian GI Kranjska Klobasa.A borderline case is Australian Fetta.With the addition of an additional t, non-Greek producers selling on the Australian market seem to be anticipating potential protection of Feta as a GI.
In spite of the limited number of names that would be at risk, a public consultation in Australia generated 400 objections 7 to the proposed EU GI list of 172 names.Although we do not know the specific GIs to which the objections were targeted, there are on average more than two objections per GI demanded-perhaps indicating overblown fears.

| Hypotheses and controls
Historically, GIs were created in Europe to protect renowned wines such as Burgundy, Champagne, Chianti and Port from cheaper local imitations and price erosion (Meloni & Swinnen, 2018).This article seeks to extend and then test this historical theory of the domestic origin of GIs to current-day GI trade policy.In particular, it is unclear to what extent the two historical conditions for the creation of GIs are relevant for the current demands of extra-EU protection.When seeking extra-EU protection, are EU demands also driven by (1) GIs being established on those markets or (2) starting to face imitations?
When operationalising 'being established', as for GIs' historical origins, the presence and success of the authentic EU GI variants is relevant.The corresponding hypothesis is: H 1 The EU is more likely to demand protection of GIs that are well established in the partner market.
When operationalising 'being imitated', identically named non-GI products from multiple origins may be relevant: from the EU outside of the protected region, from the partner country itself, or from third country producers.This is why the theory presented here is an extension of the argument in Meloni and Swinnen (2018).In the historical case, cheap local imitations led to GIs being established.In the present case of extra-EU trade agreements, identically named non-GI products could have multiple origins.The corresponding hypothesis is: H 2 The EU is more likely to demand protection of GIs that are imitated on the partner market.
Considering the literature on the protection of GIs in trade, there are other relevant factors to control for.First, GIs with higher worldwide sales are more likely to be protected in FTAs (Huysmans, 2020).This may be simply due to their economic importance and a desire to protect them worldwide, or to lobbying by producers or member states.

6
As explained in the data section, products with GI names listed as ingredients were omitted from the analysis for reasons of feasibility.
7 See https://www.dfat.gov.au/trade/agree ments/ negot iatio ns/aeuft a/publi c-objec tions -gis/Pages/ default; the number of over 400 submissions was communicated in unclassified private communication from the Australian Department of Foreign Affairs and Trade (DFAT) to the author.The content of the objections is not public.
Second, in the so-called Southern Five countries-France, Greece, Italy, Portugal and Spain-GIs and their external protection have a cultural or 'gastronationalist' importance (DeSoucey, 2010;Huysmans, 2020;Huysmans & Swinnen, 2019).This makes their GIs substantially more likely to be protected in FTAs.A potential mechanism is press coverage: although protection of GIs may lead to positive coverage of an FTA, the lack of it may attract negative press (Beattie, 2019;Christides, 2013;Malkoutzis, 2016;Reuters, 2018).
Third, GIs that have been registered in the EU for a longer time are more likely to cover important products, which are also more likely to be on the radar of EU policy-makers.In addition, the year of registration may affect the demand for protection if length of within-EU protection plays a role.
Fourth, there may be path dependency as well as residual unobserved heterogeneity at the GI level.To mitigate the risk of protracted negotiations, the EU may as a rule seek protection in all markets for a shortlist of GIs that are only relevant in some of them.In addition to controlling for this causal mechanism, including whether a GI has been protected in previous FTAs may also reduce the omitted variable bias due to unobserved heterogeneity at the GI level.
Finally, there may be a difference between protected geographical indications (PGIs) and protected designations of origin (PDOs).As the latter require all production steps to take place in the defined area, they may be more exclusive and more valuable to protect.

| Data
Basic data on EU GIs registered by the end of 2017 were collected from the DOOR database, which has now been replaced by eAmbrosia. 8he dependent variable Demand is 1 if the EU asked Australia to protect the GI and 0 otherwise.It was newly coded from public negotiation documents.9All food categories combined, the EU asked for protection of 172 GIs out of 1337 registered by the end of 2017.For cheese (class 1.3 in the EU GI system), 56 GIs out of 234 were demanded.For processed meats (class 1.2 in the EU GI system), 31 of 174 were demanded.Together, these two categories cover over half of EU demands: 87 of 172.
Focusing on cheese and processed meat GIs, data were collected on the availability of EU GI products or identically named non-GI products in the leading Australian supermarket, Woolworths, in 2020.These data were collected from the website,10 by the author and a team of research assistants.
Two supermarket chains dominate Australian retail: Woolworths (independent from the UK Woolworths Group) and Coles.Both have about 30% market share, but Woolworths accounts for over half of the online grocery market (Mitchell, 2020).Hence the focus on Woolworths.
Woolworths is Australia's largest supermarket chain, operating 987 supermarkets according to its 2020 annual report.It is not a hard discounter, and offers both national brands as well as private label products.The stores are centred on groceries but also offer non-food products.Given that Woolworths is Australia's largest chain of supermarkets, their product range is arguably representative of general Australian retail.Of course, specialty shops for cheese or delis may have a wider offer of GI products, but they are visited by less consumers and represent smaller volumes.
The main explanatory variable corresponding to H1, Woolworths, is a dummy variable equal to 1 if an authentic EU GI product was sold in Woolworths in 2020.This variable is a proxy for how established an EU GI is in Australia more generally.If a GI is sold at the most popular supermarket chain of Australia, presumably it is sold elsewhere in Australia as well.For cheese, 11 out of 234 different EU GI cheeses were found at least once.EU GI products were identified from the inclusion of PDO or PGI in the name or description, the country of origin, and the presence of the relevant EU logos shown in the Appendix S1.
For tractability, data were collected only on products in their pure form and not as ingredients listed either in the product name or the list of ingredients.For instance, downstream products such as 'Parmesan crackers', where not coded as containing either authentic GIs or identically named non-GI products.Although it would be very difficult to identify whether EU GI products were used, the use of GI names for non-GI ingredients would also be banned if protection is granted; see European Commission guidelines 2010/C 341/03 of 16 December 2010.The omission of GI names as ingredients is hence a limitation of this analysis.
Woolworths was complemented with 2019 trade data at the CN8 level, collected from Eurostat Easy Comext.For cheese, the mapping by Curzi and Huysmans (2020) of GIs to CN8 trade codes was used.As they explain and document, for cheese this mapping is fairly detailed, containing separate codes for Roquefort, Feta or Gouda for instance.For processed meats, the CN8 trade codes provide much less detail, grouping all kinds of dried ham, for instance, in code 02101131.
The variable sales AUS is the logarithm of estimated GI exports to Australia in 2019.It is the second variable corresponding to H1: the higher sales to Australia, the more established the GI.To take into account the presence of zeroes, the logarithm was taken of sales in euros plus one.Following the approach of Huysmans (2020), to estimate GI sales, the exports from the GI country in the relevant CN8 code are divided by the number of GIs from that country in that trade code.As discussed in Huysmans (2020), this approach results in a rough proxy, since sales of non-GI products are also included.Unfortunately, trade data are the best systematic data available.
A third complementary variable for H1 is Unit Value.This is the price per kilo of exports to Australia in 2019 in the relevant CN8 code, calculated using export value in euros and export volume in kilograms.Unit Value is a proxy for how established an EU GI is, although it may also capture other factors like higher quality or market power.The more authentic EU GI products are being sold rather than generic variants, the higher the average price per kilo in a CN8 code is expected to be.For example, the more actual Greek Feta is exported in CN8 code 04069032, and the less generic white cheese, the higher the average unit value in that trade code will be.
Corresponding to H2, Imitated is 1 if an identically named non-GI product was sold in Woolworths in 2020.For cheese, identically named non-GI products were found for 3 of 234 GIs.Note that this includes identically named non-GI products originating in the EU, in Australia, or in third countries.The variable codes non-GI products that would likely be banned if Australia would start protecting the GI names on the terms requested by the EU.For example, the EU typically does not seek protection of partial names such as 'Gouda'.Only the PGI 'Gouda Holland' and the PDO 'Gouda North-Holland' are protected in the EU, and only for the former has the EU sought protection in Australia. 11However, it does seek protection for 11 On partial names within the EU, see also footnote 2 on Holsteiner Tilsiter and Orkney Scottish Island Cheddar.On partial names in Australia, the list of EU demands states 'For greater clarity, the EU has confirmed examples of where the protection it is seeking would not extend to the use of parts of EU GI names.This is identified in the list below by text that has been underlined.For example, for the EU GI "Camembert de Normandie", protection is not sought for the name "camembert" when used by itself.In the EU, use of underlined names is permitted, as long as they are not used in a way that may deceive or mislead consumers as to the true origin or quality of the product.' translations of GIs.No official list of translations is provided; all customary translations were checked in the coding process.Arguably the most controversial recognised translation within the EU is 'Parmesan' for Parmigiano Reggiano; although it is not a literal translation, the evocative power is deemed such that the term Parmesan should not be allowed as a generic descriptor of cheese.
Only four cheese and processed meat GIs were found to be imitated: Grana Padano PDO with a product called 'Parmesan Grana Padano', Kranjska Klobasa PGI with 'Kransky sausages', Parmigiano Reggiano PDO with 'Parmigianno [sic] Reggiano' and several products called 'Parmesan', and Feta PDO with multiple products called 'Fetta'.Two Swiss Gruyère products were for sale, but also in the EU the homonymous Swiss GI is recognised in parallel with the GI for French Gruyère; see implementing regulation EU 110/2013 for details.
A set of variables controls for the relevant factors described in the previous section.The first control variable, sales, covers estimated worldwide exports of the GI using the same approach to construct sales AUS.Southern5 is a dummy equal to 1 for GIs from France, Greece, Italy, Portugal and Spain.YearReg codes the year a GI was registered in the EU; its food GI system started in 1996 and the number of GIs continues to grow.Listed before counts the number of previous FTAs out of a list of 11 where a GI was protected (Huysmans, 2020).The dummy PDO is 1 for PDOs and 0 for PGIs.
Table 1 gives an overview and descriptive statistics of the main variables for cheese GIs.The Appendix S1 provides a correlation matrix.Because the CN8 classification is less detailed for meat products, the variables sales AUS, Unit Value, and sales are less accurate for meat products.Hence the main analysis focuses on cheese GIs only.

| Methods
Since the dependent variable is binary, probit regressions will be used.As it turns out for H2, Imitated = 1 perfectly predicts being demanded for protection: the three imitated cheeses were all demanded by the EU.Although this is consistent with H2, it means no regression coefficients can be estimated for this variable.A qualitative analysis will discuss identically named non-GI products instead.Omitting Imitated, probit regressions of the following form will be estimated: where coefficients 1 to 3 correspond to H1 and X is a vector of control variables.The three variables of interest are three different proxies for one common underlying factor: a GI being well established on the Australian market.
To the likely limited extent that EU demands in 2019 could lead Woolworths to start stocking authentic GI products already in 2020, there could be reverse causality from Demand to Woolworths.However, if Woolworths would be willing and able to respond so quickly to new information, then it seems they could just as well wait for the negotiations to end to observe the final list of GIs that Australia would start protecting once the FTA goes into force.In addition, the 400 objections made to EU demands show that Australian producers and retailers are reluctant to accept EU demands regarding GIs, making it very unlikely that they would respond proactively to demands not yet granted by the Australian government.To the contrary, one would expect them to drag their feet until the FTA came into force.Finally, reverse causality seems even less likely for the 2019 Sales AUS and Unit Value data, because those are contemporaneous to the 2019 demands.
Because only 56 out of 234 GIs are demanded for protection, a penalised likelihood regression (Firth Logit) is also reported.This is similar to rare events logit analysis (see Firth, 1993;King & Zeng, 2001).

| Quantitative analyses
Table 2 reports the main results, which are in line with H1: the EU is more likely to ask for protection of GIs that are well established in Australia.
Model 1 is a univariate regression using the main explanatory variable Woolworths.Model 2 adds the two other variables for H1.Model 3 adds the controls.In all three models Woolworths is strongly statistically significant, corroborating H1.The two alternative variables for H1, sales AUS and Unit Value are not very significant.Perhaps sales AUS is too rough a measure even for cheese GIs, given that no specific GI sales are recorded and it is only an estimate using trade data.In Model 3, Unit Value is positive and significant at 10%, indicating that there is some evidence for protection being demanded for GIs with high unit values on the partner market.The control variables all have the expected signs, except for the registration year of the GI.
Model 4 uses a penalised likelihood logistic regression (Firth, 1993).The results are fairly similar.H1 is still supported, with Unit Value being slightly more significant than in the estimation of Model 3. Note that the coefficient magnitudes are not directly comparable because of the logit link function rather than the probit link function being used.Using a rare events logit regression (King & Zeng, 2001) gives similar results.
Putting the control variable for worldwide sales on the x-axis, Figure 1 plots the predicted probability of Demand based on Model 3, with 95% confidence intervals.It shows that the effect of being present in Woolworths is a substantial predictor of demand for protection.The Appendix S1 shows a similar plot, with Unit Value on the x-axis.
Also based on Model 3, 89% of observations are correctly predicted, versus 76% with an empty model.Although this fit is fairly high, the model does still incorrectly predict that protection would not be demanded for 24 GIs.This means that, like any model, the model does not seem to fully capture all drivers of the demand for protection.Further details on model fit are reported in the Appendix S1.
Models 5 to 8 reported in Table 3 increase the scope to include processed meat GIs as well as cheese.A dummy variable Cheese was added here.It controls for whether protection is more likely to be demanded for cheese GIs versus meat GIs.
For Models 5-6, the variables sales AUS, Unit Value, and Sales all make use of a mapping of cheese GIs to CN8 trade codes.For cheese, the CN8 classification is fairly detailed.For instance, Roquefort has its own trade code 04064010.The codes for meat products and other GIs are much more general.For instance, code 02101131 covers 'Domestic Swine Hams and Cuts Thereof, Dried or Smoked, With Bone in', which for Italy alone combines together 10 products such as Prosciutto di Parma PDO or Prosciutto di San Daniele PDO.
In addition to the problems of classifying meat at the CN8 level, some cheese and meat GIs may be sold in multiple CN8 codes depending on how aged a cheese is or which exact cut of meat was used (e.g., a hind leg or a shoulder).To deal with this problem, Models 7-8 use a less detailed classification into Harmonized System 6-digit codes (HS6).These data are less precise but the risk of mis-or double classification is smaller.The mapping of GIs to HS6 codes was taken from Raimondi et al. (2020) and updated for GIs not in their sample.
Because the data for sales AUS, Unit Value, and Sales are measured at the country-category level, Model 6 and Model 8 use country-category clusters for the standard errors rather than regular robust errors.Model 6 uses 138 CN8-country clusters and Model 8 uses 73 HS6country clusters.The results for the variables of interest are largely similar to those for cheese above.The control variables YearReg and PDO now become significant in some regressions, while worldwide sales, surprisingly, loses significance.
Across Models 5-8, the results for the main variable of interest, Woolworths, are similar to Model 3 though slightly smaller in magnitude and less significant.Unit Value is strongly significant at the CN8 level, but not at the HS6 level-consistent with the CN8 classification being more fine-grained and hence having less measurement error.Among the control variables, Southern5 is surprisingly negative and significant in these models.In models with standard errors clustered at the country level as in Huysmans (2020), it was not significant.The discrepancy may be explained by a difference in the dependent variable: whereas Huysmans (2020) looks at GIs that are actually protected in finalised FTAs, the current analysis considers demand for protection instead.The Southern Five countries may be especially persistent in not allowing their requested names to be dropped over the course of negotiations.
As a further robustness check, the Appendix S1 reports regressions using the count of products per EU GI in Woolworths, rather than a dummy for presence.The results are consistent with the main analysis: protection is more likely to be demanded for products with a higher count.
Finally, the Appendix S1 reports a robustness check using absolute estimated sales in million euro rather than the logarithm.As in Model 3, Woolworths and Unit Value have positive and significant coefficients.

| Qualitative results
As stated in the data and methods sections, only four cheese and processed meat GIs had identically named non-GI products on sale in Woolworths in 2020: Feta, Grana Padano, Kranjska Klobasa, and Parmigiano Reggiano.
Feta is a controversial GI even within the EU.It was first registered in 1996, cancelled in 1999, and reinstated in 2002 (Evans & Blakeney, 2006;Gangjee, 2007).As opponents of the Feta GI and GIs more generally like to point out, Feta is not a place name (Beattie, 2019;Gangjee, 2007).Nonetheless, the Commission has announced it is taking Denmark to court over its extra-EU exports of non-GI 'Feta' (Wax, 2019). 12In the EU, Feta can only be produced on the Greek mainland or the island of Lesbos, and in accordance with the product specification.For instance, at least 70% sheep milk has to be used, with goat milk to cover the remainder.
Perhaps in anticipation of a future agreement protecting the name Feta, the identically named non-GI products of Feta found at Woolworths were all spelled as Fetta, with double t.Four of these were from Denmark.On their website, 13 Australian producer Olympus Cheese offers the following explanation for their use of the name Fetta: 'Being cheesemakers of both Greek and Cypriot descent ourselves, we are quite mindful of the process which came about to obtain Protected Designation of Origin (PDO) status for Feta.[…] Olympus Cheese proudly makes a "fetta" cheese, but you will always note the distinct double "t".[…] Sheep's milk and goat's milk are not abundant in Queensland on the scale of our production.Therefore our fetta cheese is made on [sic] cow's milk or buffalo milk.' It is doubtful whether the additional t in Fetta will be enough to be compliant should Feta be protected in Australia. 14On the EU market, US producer Kraft sold a cheese they called 'Parmesello', which is reminiscent of the PDO cheese Parmigiano Reggiano or Parmesan in translation.They were forced to rename it to 'Pamesello', because Parmesello was deemed too close to the protected GI (Babcock and Clemens, 2004).A compromise may be possible for Feta.In CETA with Canada it was partially protected; existing Canadian producers were grandfathered, allowing them to continue production and sales in Canada (Huysmans, 2020).
Grana Padano is also a controversial cheese name.Although the EU typically seeks protection only for compound names, this does not hold for Grana Padano: the name Grana in isolation is also asked to be protected.In fact, on 31 July 2019, the Grana Padano Cheese Protection Consortium sent a public letter to this effect to the Australian Department of Foreign Affairs and Trade.The letter does leave an opening for grandfathering pre-existing Australian users of the term Grana.
The case of Kranjska Klobasa is one of a translated GI: Klobasa means sausage in Slovenian, and Kransky is a customary translation of the Kranjska region, sometimes also translated as Carnolia. 15The product for sale at Woolworths even featured a Slovenian flag, although it was 'made in Australia from at least 40% Australian ingredients'.Around the flag is the text 'inspired by the flavours of Slovenia', which arguably makes clear the product is not actually Slovenian.Yet under the standard of protection sought by the EU, such a disclaimer would not suffice to avoid breaching the rules.
For Parmigiano Reggiano, many of the identically named non-GI products were in fact called Parmesan.As pointed out earlier, within the EU Parmesan is considered a translation or evocation of Parmigiano Reggiano.However, in past trade agreements, exception has been made to allow the name Parmesan to be used.For instance, Annex 20-B of CETA exempts Parmesan and a number of other translated names like Black Forest Ham and Bavarian Beer.
To conclude, the qualitative study revealed only four identically named non-GI cheese and processed meat products being sold at Woolworths, which strongly suggests that relatively few imitations are being sold in Australia.In other words: very few products in Australia seem at risk of having to change names should EU demands for GI protection be granted.

| CONC LUSION
Combining the quantitative and qualitative results, it appears that the main goal of EU demands regarding GI protection in Australia is to foreclose future imitations rather than to prevent current uses of identically named non-GI products.This EU strategy may be worth it if changing tastes or a trade agreement would lead to greater popularity of EU GI products and hence a stronger temptation to imitate them in the future.
A policy implication of this research is that the EU could be more explicit on wanting to protect EU GIs that are well established in partner countries, rather than seeking to claw back names that have long been generic or imitated under the GI name.An ad-hoc example of such explicit clarification is the 2019 agreement between the Consorzio Tutela Mozzarella di Bufala Campana, the US Dairy Export Council (USDEC) and the CCFN.This agreement clarifies among all parties that the term 'Mozzarella' in isolation is generic, while providing 'greater support for robust protection in the United States and around the world for the Mozzarella di Bufala Campana PDO' (USDEC, 2019). 16his article's findings about the drivers of the demand for GI protection mesh well with research on the outcomes of protection.Curzi and Huysmans (2021) find no strong acrossthe-board effect of protection on exports, suggesting that limited displacement of identically named non-GI products is taking place.Potential GI export boosts, it seems, are more likely to come from fewer future imitations.Looking in more detail, Curzi and Huysmans (2021) do find an export effect of the protection of GIs with already high market shares in export markets.This detailed finding also meshes with the demand for protection focusing on GI products that are already well established: for these products, it seems, exports do increase due to GI protection.
In terms of future research, if the EU and Australia conclude an agreement, the newly collected data presented in this article can be used to examine which GIs were demanded by the EU but in the end not granted by Australia.In particular, contrasting the findings of the present analysis to Huysmans (2020) suggests that the Southern Five (France, Italy, Greece, Portugal, Spain) have been particularly successful at making sure that their requested names are not dropped, but protected in the final agreement.Future research should also expand the scope beyond cheese and processed meat GIs, although these are two large and contentious categories, and although trade data is less suitable to analyse categories other than cheese.Finally, future research could consider downstream products that list GI-named products as ingredients.
To conclude, this paper has shown that EU demands for external protection of GIs are focused on GIs where the original EU products are well established in trading partners, rather than at clawing back GIs from widespread current generic use.This finding is especially relevant in light of ongoing EU-Australia and EU-New Zealand negotiations for a trade agreement and the EU-US 'war on terroir'.Evidence presented in the article suggests overblown fears in Australia and the US of the protection of GIs and the need for better explanation of the EU's objectives.

AC K NOW L E DGE M E N T S
The author gratefully acknowledges research assistance by Josephine Holland Moritz, Marianne Kramer, and Marie Pöhlmann.For comments on previous versions, I thank Gero Laurenz Höhn, Paula Zito, Justin Hughes, anonymous reviewers, and the editor.The paper has benefited from presentations at Utrecht PPE, KU Leuven LICOS, Utrecht USE, AES 2021, ICAE 2021, and ETSG 2021.All shortcomings remain my own.Food Policy, 46, 22-36. Agri, D.G. (2012).Working document on international protection of EU Geographical Indications: objectives, outcome and challenges.AND-International (2019)