Knowing Where Organic Markets Move Next – An Analysis of Developing Countries in the Pineapple Market

As consumers’ demand for organic products grows, selling organic products potentially opens up profitable market participation options for farmers in developing countries. This paper studies two aspects of profitability for the producers. It uses hedonic demand theory and empirical analysis to examine the relation between conventional and organic markets using the strongly growing pineapple market as an example. This analysis confirms a nonlinear dependence of the organic market on the conventional one and a non-declining premium. The author concludes that there is a larger potential of the organic market and hence the number of farmers in developing countries who can potentially benefit from growing organic products. JEL L11 O13 Q13 Q17


Introduction
Organic market growth rates are around 10%, far higher than those of conventional markets and supermarkets have started offering organic food as part of their usual range of products.
Consumer demand for organic products is concentrated in North America and Europe; these two regions comprise 97% of global revenues (Willer and Kilcher, 2009). Organically grown pineapple has also become more popular among consumers. Like other tropical fruit, it is grown almost exclusively in developing countries and like other organic products, organic pineapple earns a premium price on the market compared to conventional varieties. Hence, the shift from conventional to organic production might be an opportunity for small and middle-sized farmers to reap higher returns from their investments. Since this change, however, requires costly adjustments of production techniques as well as considerable costs for certification, several aspects of organic production need to be considered when trying to determine its profitability. Another important aspect of profitability that has been disregarded in the previous literature so far is the relation between the organic market and the conventional one and its likely future development. Besides a price premium for the organic product this includes the co-movement of the two prices. In this paper we restrict our focus to this price dimension of the profitability of organic production.
The willingness to pay (WTP) a higher price for organic food based on perceived desirable characteristics has been well-documented. The academic literature has shown the existence of a, quite variable, price premium for organic food products (e.g. Boland and Schroeder, 2002;Huang, 1996;Loureiro and Hine, 2002;Thompson, 1998). We take a different approach and deduct dynamic characteristics of the demand functions from price behavior over time. Thereby we are able to provide more general results than by using survey based methods that use cross-section data based on choice experiments rather than on actual buying behavior over time (Huang and Lin, 2007 is an exception). Although our method is indirect it has the advantage of measuring what consumers are actually buying and paying in the marketplace when they have a choice between organic and conventional produce. Despite its importance for the further promotion of organic certification in developing countries, this has not been studied before.
Applying state of the art time series methods, we analyze spatial price transmission between conventional and organic pineapple on the European market by looking at prices for pineapple from Africa and Latin America respectively. Our observations not only confirm the existence of a non-declining price premium for organic products, the analysis also shows that the conventional market seems to act as a price leader for the organic market while being unaffected by organic price behavior. However, organic prices do not follow conventional prices one by one. Our results show the existence of lags and thresholds below which organic prices are unaffected by conventional price changes. These thresholds and the corresponding price adjustment behavior do not change over time, even while the organic niche market expands. Theoretically, this observation can be explained when the core demand for organic products expands faster than supply. Hence, one important implication of our analysis is the potential for the scalability of the organic market.
The rest of this paper is organized as follows. First, an introduction to the market for pineapple is given. Then, a theoretical background for the study is presented. Afterwards, the price data for conventional and organic pineapple is described and spatial price transmission between the organic and conventional markets is analyzed using time series techniques such as co-integration and vector error correction models. The paper ends with a conclusion.

The Market for Pineapple
Pineapple is well suited for this analysis because it is a relatively homogeneous good, compared to, for instance coffee, where a lot of different varieties and quality grades prevail.
This homogeneity is relevant in trade and exists because it is difficult to control for quality of single pineapple at low transaction costs. In the definition of Nelson (1970) Fold and Gough, 2008).
In the early 2000s, the wave swept to Europe. The resulting brisk upward trend in MD2 pineapple supply induced a price fall for the MD2 variety (Faure et al., 2009). By today, the price premium on MD2 which was up to 100% at market entry is almost non-existent. The formerly dominant variety, Smooth Cayenne lost market share from over 90% at the end of the 1980s to almost nonexistence today (Loeillet, 2004). The MD2-variety has become the standard variety consumed in the EU.
The most globally traded conventional fresh tropical fruits (bananas and pineapples) are primarily produce in large-scale plantations owned by transnational companies who also engage in contractual arrangements with local producers. A few large multinational companies mostly control the supply of pineapples to the large retailers within a tightly structured supply chain. This might lead to high entry barriers for small farmer market participation as indicated by many researchers (e.g. Minten et al., 2009). By contrast, organic produce is mostly produced by smallholders and does not yet rely as much on vertically integrated supply chains. For developing countries with a significant share of smallholders in production such as Ghana, the support for diversification of exports towards niche markets (for example organic markets) could therefore increase the profitability of production. In niche markets, which tend to be smaller by definition, farmers can exercise more bargaining power whilst at the same time meeting the latest requirements on quality, traceability, packaging, and standards such as GLOBALGAP 3 or organic might hold the key to good profits (Minot and Ngigi, 2004 Willer et al., 2008). In the EU, it is now estimated between 2.5 and 4.5% of total food sales.
For organic pineapples market growth has been even larger. It is assumed that the permission to use ethylene for flower induction in organic production in 2005 played an important role in the high growth rates in the organic pineapple market. Taken as a whole, Europe is the largest market for organic products, and although available data is very imprecise and often out-dated, it is assumed that this holds also for the organic pineapple market. According to estimations by the Sustainable Markets Intelligence Centre (CIMS), the European market for organic pineapple was about five times the size of the US market in However, not only the growing demand makes organic cultivation attractive for producers. Some studies explain the growing interest in organic agriculture in developing countries also by the fact that it requires less financial input and places more reliance on the natural and human resources available (Willer et al., 2008 amongst others). Hence, it is worthwhile to analyse if switching from conventional to organic production might indeed result in higher profits for farmers. As a starting point, integration of the two markets is evaluated by looking at the price developments for organic compared to conventional pineapple.

Theoretical Background
Consumers who buy organic products do so because of their perceived superior attributes.
Hedonic demand theory can help to formalize the relation between conventional and organic prices in order to provide an analytical framework for the interpretation of empirical results.
The hedonic approach disaggregates commodities into characteristics and estimates implicit values for units of the characteristics. The hedonic price function specifies how the market price ( of the commodity varies as its characteristics (z) vary (Ladd and Suvannunt, 1976). The simple assumption behind this theory is that utility is derived from the properties or characteristics of goods. We focus on one attribute of interest only, the organic nature of a product which is assumed to be otherwise homogeneous.
Standard maximization of a consumers' utility function U(z, x; ) subject to a budget constraint, where x is the commodity, and is a vector of parameters characterizing the individual consumer, gives rise to a vector of demand functions for the characteristics of the good: , ; , ; , ; (3.1) denotes the vector of first derivatives of a hedonic price function with respect to its arguments, i.e. the vector of implicit prices of each property. If the distribution of and z is known, then the hedonic price function can also be written as a function of these arguments, and hence the price function depends on the parameters that characterize the distribution of preferences and supply (Epple, 1987) Our case is a simple hedonic model, where the number of characteristics is fixed and z has only two values; let z = 1 if a product is organic and z = 0 otherwise. We add a time dimension in which the price when z=1 in time t depends on past prices of the good in both states (organic and conventional) and other hedonic characteristics of the good. We assume that the other hedonic characteristics are time invariant. Hence if organic pineapple is on average yellower from the outside in time t=1, we assume that this is also the case in all other periods. In addition, if information is imperfect, rational consumers gather information about a characteristic if the marginal cost of obtaining the information is smaller than or equal to the marginal utility it generates (Combris et al., 1997). For most consumers, it is not easy to judge the taste from the outside of a pineapple. Accordingly consumers may decide to make their choice primarily on the basis of the easily accessible characteristics, for instance size and certification status. This limits the number of relevant characteristics. Hence, if the status of z is valuable and easy to assess, ignoring other product characteristics may not be a problem. These simplifications make it easier to estimate the value of the organic attribute, which can then be approximated by the price difference between organic and conventional pineapple. Furthermore, we ignore the household budget constraint because, by focusing on the organic pineapple price premium, we touch such a tiny part of the overall household budget that we can safely assume the constraint to be non-binding. Hence, we refer to the case in which households have identical incomes and characteristics, and different tastes.
We do not estimate a (reduced form) hedonic model, but use it to understand the empirical results from the estimation of the dynamic relationship between the conventional and organic prices. For this purpose we derive a number of hypotheses from the above described hedonic price theory that can be investigated with our price transmission analysis.

Hypothesis 1: The organic price moves along with the conventional price, but with a lag.
This phenomenon can be explained with imperfect information. In Rosen's original framework, consumers and producers make their decisions on the basis of perfect information. This assumption is in reality often not met. In our simple example the consumer might not observe the prices for z=1 and z=0 at the same time and might consider it too costly to look for the reference product in another shop as long as the price stays within a certain range that is perceived as "normal". On the other side, assuming that the wholesaler estimates the size of the WTP for an organic premium, he will use the conventional prices as reference. But he might only have knowledge about yesterday's pineapple prices not about pineapple sold at the same time. FOB (free on board) prices may also be pre-fixed with the supplier for a certain shipload (which takes between 10 and 15 days). These two considerations would lead to lags in the dynamic relationship between the observed prices.
Hypothesis 2: Cross-price elasticities are low within a certain range of price changes, and high when crossing a certain threshold.
This can be represented by two related demand curves that are connected by cross-price elasticities. Imagine the price for the good where z=0 falls, while the price stays constant for z=1. Then we assume that there is a tolerance range in which consumers do not react to this price change. This range exists due to imperfect information about the price difference between the two regimes and sluggish demand response which can be explained by habits.
Since pineapple is a perishable non-staple food product, small price ranges will not switch, postpone or anticipate buying decisions. This causes low cross-price elasticities within this tolerance range of price changes and considerably higher ones when crossing the tolerance threshold. This threshold cannot be expected to be the same for all consumers, but again falls within a certain range, and hence a (fuzzy) jump in the elasticity is expected. Because markets for perishable products have to adjust fast to changes, this hypothesis should be reflected in prices changes.

Hypothesis 3: The organic premium and hence the WTP for organic products depends on the relative size of the two markets in a non-linear way.
When the organic market is expanding at a different speed than the conventional market, the premium is likely not constant over time. The demand curves shift with changing consumer preferences. The supply curves move to the right as more farmers start to produce pineapple, and the movements of the curves are interrelated, but not perfectly collinear.
Changes in preferences affect both demand curves, but the size and timing of the effect may differ. We expect the demand for organic pineapple to shift faster than the demand for conventional pineapple, since the former market is in an earlier stage of the product life cycle. This may trigger several countervailing effects.
On the one hand, the WTP for the organic attribute may decrease when the size difference between the two markets decreases. This would be in line with observations in marketing research, that the price difference between a standard product and a specialty product decreases when the latter becomes less rare, and therefore less special. This also makes sense when we separate the hedonic demand into consumer groups with different marginal monetary values of the organic characteristic (Ladd and Suvannunt, 1976) and assume that the relative WTP between groups is constant. The first consumer group that buys organic products is the one with the highest WTP, the second group has the second highest WTP, and so on. When the market grows beyond the core market (the first consumer group), it can do so only by expanding into consumer groups with lower WTP for organic.
Hence, as the organic pineapple market expands, prices for organic pineapple might drop.
On the supply side economies of scale in production, transport (which are included and comprise up to 50% of import prices), distribution and marketing could also lead to decreasing premia due to decreasing costs that affect the supply curve.
On the other hand, if consumer preferences for organic expand fast enough, they might absorb the increasing supply. When the core market for the organic attribute increases against an inelastic short run supply, the premium rises. In the longer run more producers can start producing organically and the premium will be adjusted downwards. Since conversion to organic takes several years, where preferences can change very rapidly, shifts in the supply curve occur much slower than they may in the demand curve. 6 In sum, we can derive information about the hedonic demand forces at work by studying the transmission between organic and conventional prices over time. The interaction between demand and supply for the organic attribute will determine the development of the organic relative to the conventional price. We have described three different effects: lagged response, a threshold effect, and demand and supply shifts.

Prices for Conventional Pineapple
Average monthly wholesale market prices in € per kg from Europe 7 are used in our empirical analysis. As data on organic pineapple prices are neither publicly recorded, nor readily available from the parties involved in the trade, the data collection process was tedious, and we had to use a number of data sources. The data is taken from International Trade Centre's market news service and from several European fruit trading companies. We distinguish between organic and conventional and focus on sea transported pineapple, hence exclude air transported pineapple 8 . We limit ourselves to the currently dominant MD2 variety. By doing so, we deliberately exclude a number of hedonic characteristics (such as the variety) that might otherwise bias our results.
The data could be obtained from the two dominant regions of origin for fresh and dried pineapple in Europe, Latin America (in our dataset -as in reality -mainly Costa Rica and less dominant Ecuador) and West Africa (Côte d'Ivoire, Ghana and Togo). Due to severe gaps in the data for single destination countries, the monthly prices for conventional pineapple were averaged over all destination countries for each of the two regions of origin.
Through this averaging, a conventional time series over the period January 2001 to July 2011 could be obtained. The data for organic pineapple prices covers the period September 2007 to August 2011. In this section, the time series for organic and conventional prices is analyzed using descriptive and graphical methods separately and jointly. Whenever we examine both prices jointly, we restrict ourselves to the shorter period (2007 -2011).
Nevertheless showing the longer time series for conventional pineapple allows us to explain some general trends.
The evolution of prices over the last 10 years for conventional pineapple from the three sample countries is shown in Figures 3.1-3.3. There is a general trend towards lower pineapple prices observed in the market. The widening gap between volumes and values of EU pineapple imports in Figure 3.1 makes the fall in prices in general for pineapple clear.
Whereas the volume of pineapple imports has more than doubled since 2003, the value of pineapple imports has increased only by about 50%.
We then look at the prices in more detail. Rica had a first mover advantage.

Organic Premia
Organic certification is a value-addition method. In fact, organic products are usually sold at significantly higher prices than conventional products. According to CBI (2008) organic products generally fetch price premia of between 15 and 25% and numerous scientific studies have also shown the existence of price premiums for organic products (e.g. Teisl et al., 2002;Nimon and Beghin, 1999;Bjorner et al., 2004).
With regard to the potential benefits of organic farming for producers, an important question is if such price premia can be sustained in the long run or if they will vanish, as in the case of the MD2 variety. The recent developments in typical agricultural commodities like wheat or milk show that the price premium for organic products seems to be relatively constant 10 . Whether this is a temporary development or a long-term trend depends on changes in supply characteristics and in consumers' perception about the value added by the organic certification label (hypothesis 3).
The data shows that, for the period from September 2007 to July 2011, price premia have fluctuated between €0.14 and €1.02 with mean (standard deviation) of €0.51 (0.20) respectively on average (Figure 3.4) 11 . A declining trend cannot be observed over this period.
This might tell us which forces are at work with respect to hypothesis 3 12 . The comparison of the price behavior in Figure 3.4 also shows that the premium is far from stable over the observed time period. Obviously the two curves are interdependent. In this context we should take note of a particularity of the pineapple market. The supply of conventional pineapple is highly dependent on harvests in Latin America, especially in Costa Rica (see section 2 above), whereas organic pineapples are reported to come from a variety of source countries.
Hence, for instance weather conditions or new plant diseases in Latin America would influence the two markets differently. This is unobservable without information about such supply shocks. However apart from this, there are potentially market inherent explanations for these fluctuations, which will be studied in the next section, the econometric study of price transmission.

Econometric Analysis of Spatial Price Transmission
The notion of price transmission is used in different contexts in the literature. First of all, some authors test for price transmission within the value chain of a product. For example, it is analyzed if the world market price of a commodity is transmitted to domestic producers.
Other authors are interested in the difference of prices between different markets within one country, the so-called spatial price transmission. In this paper however, we study spatial price transmission between the markets for organic and conventional pineapple from Latin America and Africa in the European market. We do not use panel data methods, since there are only two regions for which data are available, which can arguably be hardly called a panel. As a result there is no information loss from analyzing the two regions separately.
We test the hypothesis that prices in the organic market are dependent on prices in the conventional market due to its dominance in size (hypothesis 1). Secondly, we analyze if small and large price changes have different effects on the respective other price (hypothesis 2). Finally, we explore if such a possible integration between the two markets decreases or increases over time as a result of the growth of the organic market and possible supply and demand shifts (hypothesis 3).
When analyzing price transmission, different price series are usually regressed on each other in order to find a possible relationship between them. However, if the time series are non-stationary, it might be the case that a relationship is established even though the series are independent from each other as shown by Granger and Newbold (1974). In order to avoid these spurious regressions in case of non-stationarity, many authors have used cointegration techniques to study price transmission and long-run relations between different prices (for example Meyer andvon Cramon-Taubadel, 2004 andAbdulai, 2000). Rapsomanikis et al. (2003) also use cointegration methods and error-correction models, and develop a comprehensive framework to test for the price transmission between local coffee markets of Ethiopia, Rwanda and Uganda and the international market.

Unit Root Tests
As in Rapsomanikis et al.'s framework, we start our analysis by testing prices in the organic and conventional markets for unit roots. As explained above, this is important in order to avoid spurious regressions when studying spatial price transmission. The time series of the two regions of origin are tested separately.

For the individual time series unit root tests, the traditionally employed Augmented
Dickey Fuller (ADF) test has been used. However, it has recently been documented that this test performs badly in the presence of small samples as the ones used in this paper. In addition, the ADF test has low power in distinguishing highly persistent stationary processes from non-stationary processes and the power of these unit root tests diminishes as deterministic terms are added to the test regressions. Elliot, Rothenberg and Stock (1996) have proposed an alternative test that addresses the above shortcomings and that has also been used to test for unit roots in the variables. For this DF-GLS test the data is first detrended using generalized least squares. In order to employ the tests, it is necessary to determine the optimal number of lags of the prices to be included. One approach often employed is to use the Schwartz or the AIC criterion. However, as shown by Ng and Perron (2001), in the presence of large negative moving-average components of the error term, these information criteria usually choose a lag length that is too short. This in turn leads to size distortions and hence overrejection of the null hypothesis. Ng and Perron (2001) propose a modified version of the AIC (MAIC) that improves on these problems. In the analysis below both the Schwartz criterion as well as the MAIC are employed.

Analysis of Cointegration and Price Dynamics between Markets
Since both Latin American price series are integrated of order one we test for cointegration. If the linear combination of the two time series is stationary, it would describe the long-run relation between the two variables. The number of cointegrating vectors in the system is determined using the Johansen test. We consider the cases without a constant or trend and with a constant in the cointegrating relationship because the series do not exhibit an apparent trend when plotted in levels (over the period 2007 to 2011, see Figure 3.4). The results are illustrated in Table A.3.5. There is clearly one cointegrating vector. We then test for granger causality. Table A.3.6 shows that Latin American conventional prices granger cause organic prices, that is lags of conventional prices improve the forecast of organic prices but not vice versa. We expected the conventional market to act as a leader due to its dominance in size; hence this result confirms our a priori expectations. The results on cointegration mean that there exists a long-run relation between the conventional and organic Latin American pineapple prices and a linear combination of the two prices that is stationary.
For African pineapple prices, since they are stationary, we do not test for cointegration. Even though we would be able to analyze the data on African pineapple in levels, for reasons of comparability we use the same models as for Latin American pineapple.
Let p = (p c p o ) where p c and p o are the conventional and organic prices respectively.
Then there exists β such that βp is stationary. Then, the long-run relation between the two prices has to be taken into account by a cointegrated version of the VAR. Therefore, the following vector error correction model (VEC) has been applied in our analysis: (3.2) ∆ is the difference operator, c indicates a constant, p ct-i and p ot-i indicate the i th lag of p ct , and p ot ,  i describes the short-run relation among p t and the i th lag, and =β, where β is the cointegrating vector defined above and  measures the speed of adjustment of the two prices to deviations from their long-run relation. All variables are transformed into natural logarithms. In order to employ this approach, the optimal lag length for the differenced price vector has to be determined. Akaike and Schwarz's Bayesian and Hannan and Quinn information criteria were used to determine the optimal number of lags to include in the cointegrated VAR. All of them suggested that estimating the model by using one lag was optimal. Therefore, the model above with only one lag has been estimated. Results are reported in Table 3.1. The cointegration equation for Latin American prices is given by: This represents the long-run relation between the two Latin American prices.
Estimating the VEC model indicates that a price increase in the conventional market, which generates a deviation from this long-run relation between the two prices, generates a price increase in the organic market, whereas an equivalent price increase in the organic market produces no significant change in the price for conventional pineapple. We see asymmetric transmission of price changes between the two markets in the sense that organic prices do not respond in the same way to changes in conventional prices as conventional prices to changes in organic prices 14 .
Considering the short-run dynamics, suggest that organic prices are strongly influenced by conventional price movements, whereas this is not true in the opposite direction. This confirms our hypothesis 1 that the conventional market acts as a price leader for the organic one.
Although our results suggest that organic prices follow prices in the conventional market, there is no reason to believe that this relation is linear. Niche markets might change at a different speed than the main market for various reasons (see hypothesis 3). Hence, the following section investigates the possibility of a non-linear relation with a threshold autoregressive (TAR) model and thereby tests hypotheses 2 and 3.

Markets
Previous studies explained non-linearities by transaction costs of spatially separated markets for the same good (e.g. Baulch, 1997;Fafchamps, 1992;Sexton et al., 1991). Unlike in these studies, in our example transaction costs are not the result of costs and risks associated with trade between such separated markets and the speed of adjustment is not necessarily dependent on the traders' access to market information. At the wholesale level information about prices in conventional markets is readily available. And we have found out that organic prices follow the price in the main market (that is the conventional market) and not vice versa.
In our case, thresholds may exist when consumers see conventional and organic pineapple as two different products. This may happen when there is a physical separationstill a considerable part of organic pineapple is traded by way of organic specialty markets as opposed to mainstream food multinationals -or when marketing and branding efforts of companies are successful. A threshold also exists due to the switching behavior of consumers: when the price difference between the organic and the conventional pineapple increases beyond the willingness to pay for an organic pineapple, then the consumer may switch and buy a conventional pineapple instead, and vice versa.
The organic premium is not constant over time (Figure 3.2). If hypothesis 2 is correct, it is possible that due to a certain willingness to pay for organic products relative to conventional goods, organic prices only respond to movements in conventional prices when the difference between these two prices exceeds a certain threshold. On the supply side, both thresholds and non-immediate adjustment can be caused by differences in competitive structures: a small number of fiercely competing food multinationals in the conventional market versus a larger number of smaller competitors and limited possibilities consumers to compare prices in the niche market. In addition, if conventional prices vary as a result of changing supply conditions from Costa Rica, organic prices might not adjust or not as much.
The possibility of a threshold would in this case be owed to menu costs and competitive structures.
In addition, the size of thresholds themselves may vary over time with the relative WTP of consumers for organic over conventional products. As stated in hypothesis 3, the threshold may vary when cross-price elasticities change over time.
In this paper, we follow the analysis by Van Campenhout (2007) who uses a threshold autoregressive model to test for integration of several Tanzanian maize markets over time. As explained by the author, the threshold autoregressive (TAR) model can be preferred over a parity bounds model (PBM) because the TAR model allows separating the two market components of transaction costs and speed of adjustment of prices. Moreover, it allows for time-varying thresholds. To analyze possible non-linearities in the relation between organic and conventional prices, we estimate the following TAR model: where m t = p c,t -p o,t is the difference between the conventional and the organic price in period t,  t ~ N (0, ²).  in and  out measure the adjustment speed, the change in the price difference as result of the previous difference itself, within the band created by the threshold θ and outside this band respectively. If the hypothesis of a threshold was wrong, these two parameters should be the same.
It is possible that the threshold is not constant but changing over time. To incorporate this possibility, the threshold θ can be modeled as a function of time: (3.5) where t ϵ (0,T).
In addition, we will allow for a time trend in the adjustment parameters  in and  out.
These two extensions can be expressed by the following second model: To estimate these two models, the data was converted into first differences. Data in this form was stationary for all the time series. To determine the threshold parameters θ, θ o and θ T , a grid search over all possible values has been performed. Furthermore, according to the hypothesis that prices only respond if the difference between them is large enough,  in is set to zero in the analysis.
The results are shown in Table 3.2. The threshold is at 63% (Latin America) and 53% (Africa) of the average differenced price in the simple TAR model, confirming hypothesis 2.
This number is quite high, but one should remember that the price changes are rather small compared to the absolute value of the price. When including time trends, thresholds for Latin American pineapple stay the same and thresholds for African pineapple increase from 46% to 61%. On the other hand, above the thresholds, adjustment speeds ( There is also no indication that the premium on organic pineapple is bound to decrease. However, since our database covers only four years, this rather indicates that more research should be done to answer this question when more data is available than a strong rejection of the hypothesis. Still, overall these results indicate thresholds in price responses that did not change significantly over the past four years, and there is also no difference in regions of origin. These results may help farmers, traders, retailers, and agencies promoting organic certification to better understand the market and predict future price movements. The availability of more data over time will improve the results.

Conclusions
As the demand for organic products is growing, this paper has tried to shed light on the longer-term profitability of organic production. Taking hedonic demand theory as basis, we empirically analyzed spatial price transmission between organic and conventional pineapple on the world's largest organic market Europe as a case study. The analysis is set up with a development perspective since organic products in general and organic pineapple in particular are niche markets that exhibit premium prices. As a result, organic production is currently promoted as a valuable agricultural alternative for developing countries. Our results imply that the conventional market acts as a price leader for the organic one. While prices for conventional pineapple are independent of organic prices, organic price movements are responding to their conventional counterparts. However, threshold analysis indicates that organic prices only react to changes in conventional prices if these changes are sufficiently large. In addition, this threshold does not change over time. Hence, despite an expanding organic niche, market integration does not increase. Our observations also do not show an upward or downward trend for the organic price premium in the pineapple market. When there is neither more integration, nor a declining price premium to be observed, while the organic market is expanding faster than the main market, this happens, according to theory, only when the core market expands faster than supply. One important implication of this observation is the potential for the scalability of the organic market. Accordingly, these results suggest that organic production can indeed be a profitable alternative for small farmers in developing countries, and it is likely to remain so in the near future.     12.47*** 7.65*** 22.13*** 11.93*** Notes: Dependent variable is the change between two periods in the price difference between the two market prices. All models are estimated without a constant. Rho (  ) denotes the adjustment parameter on the lagged price difference expressed as the percentage of mean price in the two markets, theta ( ) is the threshold expressed again as the percentage of mean price in the two markets and t is a time trend. The TAR models are three regime symmetric models with unit root behavior imposed within the band formed by the thresholds. The thresholds are identified through a grid search over candidate thresholds with as model selection criterion the minimal sum of squared residuals. As starting values for the thresholds, at least 20% of the observations were either within or outside the band formed by the thresholds. Half-lives are expressed in months and in brackets when they are based on a coefficient that was estimated not significantly different from zero. Standard errors are in brackets. *, ** and *** denote parameter estimates significantly different from zero at the 10%, 5% and 1% significance, respectively. N is the number of observations used in the estimation. Africa (1/11) -3.617*** -4.787*** -3.501** -3.031*