geacollege Faculty of Entrepreneurship AN INVESTIGATION INTO THE PRICE TRANSMISSION BETWEEN PRODUCERS AND RETAILERS WITHIN THE UK MILK MARKET David J Stubley Harper Adams University, United Kingdom 13217600@harper-adams.ac.uk Dimitrios Paparas* Harper Adams University, United Kingdom dpaparas@harper-adams.ac.uk Ourania Tremma Harper Adams University, United Kingdom otremma@harper-adams.ac.uk Luis De Aguiar Harper Adams University, United Kingdom ldeaguiar@harper-adams.ac.uk Abstract The main aim of this article is to investigate the price transmission of milk between the producers and the retailers within the UK to understand the influence of large retailers on the market. In recent times smaller dairy farms have been forced to close down because they believe that prices are not being conveyed from retailers to producers. The research interlinks well-established econometric tests, which are frequently used within vertical price transmission research to gain an understanding of the transmission from producer to retailer. These are unit root tests, cointegration tests and causality test. The main findings were that there is a unidirectional transmission of milk prices in the UK between producers and retailers. The Granger causality test shows that causality runs from the retailer to the producer and but not from the producer to the retailer. There was a significant break in 1994, which is when the MMB disbanded and has provided a new research gap. The direction of causality means that when producers are losing out to large retailers. The ECM results indicate that the prices are slow in recovering to a new equilibrium after a shock has occurred. Research specifically on the UK milk market is limited and therefore this research is a basis for future studies, which will help policy makers when moving forward post brexit. 'Corresponding Author Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 Key Words Price transmission; retailer; producer; milk. INTRODUCTION This research is going to investigate the vertical price transmission of milk between producer prices and the retailer prices within the UK. Price transmission is the process, which measures the relationship of prices between two markets. The two types of price transmission are horizontal and vertical. Horizontal price transmission explores the relationship of prices across different markets for example across two countries. This provides a good comparison between two similar markets, which is useful for benchmarking. Vertical price transmission measures the relationship upstream and downstream within the specified market. It is used to assess where value is added within supply chains and particularly within agriculture to establish the relationship between producers and big retailers. Both horizontal and vertical price transmissions have been used effectively in econometrics to investigate the relationship of prices during fascinating and challenging financial times. Agricultural products are notoriously volatile in price and therefore econometric techniques are frequently used to investigate the relationships of prices. On average UK dairy farms are increasing in size and productivity is rising causing smaller dairy producers to struggle to cope with decreasing prices paid from retailers and increasing production costs on farm. The result of this is that many smaller producers have had to resign from milk production. The anger from producers is being directed at large retailers who hold massive bargaining power when it comes to negotiating prices. In order to protect the UK dairy industry in the future it is necessary to understand where the issues are. This dissertation will conduct a literature review, which will analyse current research in order to gain an understanding of what methods previous studies have used and what they have already found. This will establish a gap in the current literature where more research needs to be undertaken and expose which methods are successful and which methods have limitations. The methodology will then outline the details of how the research is going to be completed. From the literature review it should be possible to get a good interpretation of what econometric approaches will be useful and applicable. The methodology will outline what type of data is required and how it will be collected, what theories will be applied and what statistical tests will be used. Once the research has been undertaken the results will be displayed. Relevant comments will support the results and where applicable explanations will be provided for the findings. Lastly the research will be concluded, binding together the whole research proposal and listing any comparisons with previous research, any weaknesses and any strong findings. This will establish any areas for further research. 15 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 LITERATURE REVIEW The Application of Price Transmission Agricultural markets are one of the most common areas of study for price transmission as a result Meyer and Cramon- Taubadel (2004) investigated how effective asymmetric price transmission actually is. They used a number of methods to examine possible causes of asymmetric price transmission and they investigated the empirical tests used to determine results. They found that research struggled to combine both the theoretical and methodological aspects of asymmetric price transmission. It was established that in particular the agricultural studies, which account for large proportions of this work, failed to link theory and methods together. As a result researchers in other fields of economics often overlooked these studies. They suggest that more emphasis needs to be put on the quality of the data, the relevance of the results in relation to the external economy and explanations behind the results. This led to modern research using techniques that could prove the validity and reliability of the data. Price transmission has been studied extensively and Conforti (2004) researched claims that there were as many as six factors affecting price transmission models. These were Transport and Transaction costs, Market Power, Increasing Returns to Scale in Production, Product Homogeneity and Differentiation, Exchange Rates and Border and Domestic policies. It was believed that these factors contribute to the behaviour of a vertical or horizontal price transmission model. These factors are important to understand, as they will have an influence on the results of future studies (Conforti, 2014). The research looked to provide support and point out weaknesses of price transmission within agriculture. The research was based on a range of countries all of which have a strong basic food commodity trade improving the significance of the studies results. Due to the scale of the investigation it is difficult to generalise the results however a geographical regularity was discovered providing evidence that the price transmission model is accurate. It was also found that transmission within a domestic market is more integrated than transmission between domestic and border prices. When analysing data for a given product it is more reliable to use figures from producers, retailers and wholesalers within a domestic market instead of incorporating world prices. The last finding was that price transmission arose for products that are regulated by public intervention, for example policies. Many of the potential pitfalls highlighted in this study would not affect data, which is contained within one country. The six factors Conforti (2014) identified as price transmission influencers are reciprocated in other studies. An et al (2016) found that boarder and domestic policies were key components to the volatility of wheat and flour market prices in Ukraine. In addition to this Assefa et al (2014) explained how market power affected the asymmetry of Dutch potato prices between retailers and farmers. Farm price decreases were not fully transmitted to the retailer price however farm price increases where almost perfectly 16 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 transmitted to the retail price. The limited markets available for producers to sell their potatoes explained this. With few retailers, due to their colossal size, there isn't any alternative competition for producers to market their potatoes to. Parsley (2003) inspected the influence exchange rates have on both vertical and horizontal price transmission. The exchange rate pass-through was compared within world prices and within domestic prices. The results demonstrated that individual domestic markets are more responsive to currency rate changes, which isn't reciprocated at world price levels. This was expected as previous research from Goldberg and Knetter (1996) had mirrored these results, however this was on a larger scale and therefore more reliable. Research in to exchange rates and in particular the pass-through is imperative for policy makers within countries as decisions being made will affect the domestic market structure (Baldwin, 1988). This will therefore have an impact on the price of goods and influence the price transmission between producers and retailers. These findings are significant as it confirms that the six influencers Conforti (2014) found, hold true for a large number of markets and therefore need considering when analysing milk prices. Previous Empirical Results The nature of examining price transmission means that using secondary data is the most effective data to use (Lloyd, 2017). Slagboom et al (2016) conducted online surveys to collect primary data to explore the organic dairy industries production in comparison to conventional farmers. Using online surveys meant a large amount of surveys could be conducted however participants may not be motivated to answer appropriately and therefore the validity and reliability of the data is questionable. The study is limited because of the methods Slagboom et al (2016) used in collecting data. Similarly to this Tuckett (2012) used an interview technique to gather market information on financial markets. When using quantitative tests in research it has often been argued that this method of collecting data is unsuitable (Gray, 2013). However Tuckett (2012) found that interviews could be an effective tool for backing up finical data and give explanatory narrative to quantitative data. It has to be noted that this was based solely on one interview, which shows significant fragilities within the research. When conducting this research it will be possible to give qualitative explanations for the data by reading extensively around the topic. Having access to huge amounts of qualitative data online will give sufficient explanations to back up any findings. McLaren (2015) researched world markets and their effect on local agricultural markets asymmetry in price transmission. The consequences in local markets of a poorer price transmission could mean farmers going below the poverty line (Mosley and Suleiman, 2007). Mclaren (2015) found that where there was a bigger presence of large intermediaries, big powerful organisations like Cargill, then the asymmetry is stronger. Local agricultural markets can be harmed, particularly in poorer countries by a high degree of asymmetric price transmission when large intermediaries are present. This is 17 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 also the case in recent years for UK milk producers who have seen their payments decrease to levels below the cost of production (AHDB. 2017). Investigating the relationship between the producer and retailer has become more interesting recently as it is claimed that retailers have obtained too much buying power (Acosta and Valdes, 2014). They suggest that a lack of communication between the milk sector and government organisations has lead to insufficient policies being used within dairy markets around the world. In view of this, econometric analysis has developed so that the relationships of the price between producers and retailers can be easily studied (Hassouneh et al., 2012). As a result it is possible to understand what causes the price fluctuations. Hassouneh et al's (2012) explored techniques where co-integration and whether unit roots did or did not exist. Unit roots signify whether data is stationary, which means it is reliable and valid for testing. Co-integration examines whether the data has a long run relationship. The methods they used to test for unit roots were the ADF test (Dickey and Fuller, 1979) and the PP test (Perron, 1997). They concluded that if the data had unit roots then it was applicable to test for co-integration. If unit roots are not present then instead use the vector error correction model for co-integration, with stationary data. Hassouneh et al (2012) tested co-integration using Johansen's (1988) approach. It was concluded that if there was co-integration and therefore there was a long run relationship between the two sets of pricing, then further in-depth analysis could be applied. This includes Threshold Vector Error Correction Model and the Smooth Transition Vector Error Correction Model (Hassouneh et al., 2012). If co-integration did not exist then it should be tested using prices in first difference. Weldesenbet (2013) used this methed to test the asymmetric price transmission of liquid milk in Slovakia. There were worries over the productivity of the milk market price transmissions as the country saw a divergence of prices. Johansen's (1988) co-integration test and Granger's (1969) causality test was used. It was proved that the wholesalers and producers prices were co-integrated, as were the retailers and producers prices. The direction of causality is from the producers to the retailers and to the wholesalers, which means that if the producer price changes it effects the retailer price and the wholesaler price and therefore it was concluded that the Slovakian milk market is asymmetric. The methods used were similar to the findings of Hassouneh et al (2012) and the asymmetric results coincide with a volatile milk price seen in Slovakia. Vertical and Horizontal Price Transmission The steps used by Hassouneh et al (2012) and Weldesenbet (2013) are used in much of the contemporary research on price transmission. Bakucs et al (2012) used the same steps, Unit Root test, Co-integration and Causality test to examine the price transmission in the milk sector. The thing that separated this study was that it was one of the first journals to consider the price transmission across two countries, being Poland and Hungary. After confirming cointegration exists they found that in Poland the causality runs from the retailer to the producer however in Hungary it runs from the 18 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 producer to the retailer. In Poland the retailer price affects the producer's price whereas in Hungary, like Slovakia, the producer's price affects the retailer's price. This was explained by the high power of the dairy producers in Hungary, which does not exist in Poland. What it so effective about the methods used to analyse horizontal price transmission is that it can be applied to any country and any commodity and it is comparable, like in this instance, across two countries. By comparing two countries the study is not limited in the understanding of the speed and size of price adjustments as there is a direct comparison. The differences that arose between the countries gave evidence, which actually explained the speed and nature of price transmission. With a growing uncertainty around milk producers in Europe it is necessary to compare and contrast with similar countries in order to try and gain an understanding of the problems. Most theories suggest that the producer struggles are a result of increases in price from the retailers, which are not being transmitted down steam in the supply chain. In addition Bakucs et al (2012) also conducted the tests with structural breaks, which gives further confidence in the results that where obtained as it shows any shocks and spikes in the prices were considered. Asche et al (2007) used both vertical and horizontal techniques when examining market integration and price transmission of salmon. The usual unit root testing and cointegration tests were applied but the producer prices were from the UK and Norway and the retailer prices were from France. Having multiple countries provided the horizontal aspect of analysing price transmission. The benefits of this are that there is a direct comparison of the two producing countries and therefore the trade disputes that have arisen can be answered for and settled. The results show a high level of integration and price transmission in both UK and Norwegian Salmon. There was no competition between the two countries at producer level however having a high level of price transmission means that any restrictions or advantages across the whole Norwegian supply chain will benefit or harm the UK supply chain at the corresponding level. Therefore the effects of salmon companies in Norway becoming more international could put pressure on the price of the UK producers. While an obvious advantage is that Asche et al (2007) had access to data from 3 countries and 2 complete supply chains the data was only for a six year period. This is a relatively short-medium term period and therefore the data may not be valid in the long run. In addition the French retailer prices could not be separated for Norwegian producers or UK producers leading to more potential inaccuracies. While the horizontal and vertical approaches combined have yielded more functional results, the data restrictions appear to have a large influence on how reliable and creditable the research actually is. There is very little research about the price transition patterns of milk for the UK. Considering the recent hard times of UK milk producers it is surprising that this it has not been more thoroughly investigated. Franks and Hauser's (2012) research collected data using an online survey of UK milk producers, which it could be argued would give an imbalanced view. In addition using an on-line survey to gather data may be unreliable as only those who have a really biased viewpoint will take the time to answer it. The 19 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 need for the research was because the UK's MMB was disbanded in 1994, which left a void between milk producers and milk retailers. Franks and Hauser (2012) recognised this gap and explored two titles in relation to milk prices; "marginal value in the least remunerative use" or whether "the market had put in place some other mechanism for raising the price upwards" following the MMB collapse. They found that a better transparency of prices would result in better prices for the producer. The producers achieving the best price were the ones selling direct to processors rather than selling to one of three main farmer owned cooperatives. Conclusions found that since the break up of the MMB there could have been more done to protect the prices producers were paid for their milk. While Franks and Hauser (2012) raise some interesting points, there methods mean that only milk producers have taken part in the research. For future research it would be important to get a balanced perspective by using data from both the producers and the retailers. Despite this pitfall, there is an obvious need to look in more depth at the price being paid to producers and whether it reflects the price retailers are receiving. The only current similar research to the price transmission of milk in the UK is for other products. Sanjuan, and Dawson (2003) examined price transmission between the retailer and producer for the prices of beef, lamb and pork. The purpose of this research was to investigate the affect the BSE crisis, which occurred in 1996, had on the meat industry. The methods used to examine the price transmission were the most common, unit root test, co-integration tests and causality tests. This is the same method, which Hassouneh et al (2012) discovered to be reliable when examining price transmission. By focusing on the UK Sanjuan and Dawson (2003) could investigate three different products. This differs to Franks and Hauser (2012) who compared across countries rather across different products. Both are successful and useful for looking at the explanation behind price transmission rather than just the theory. Sanjuan and Dawson (2003) found that the BSE crisis did not have any significant affect on the lamb or pork market. However there was a structural break in producer and retail prices of beef in 1996 in which the price transmission from producer to retailers was poor increasing the retailer's margin and benefitting them as a result. This is as expected, and is consistent with research that finds powerful retailers and intermediaries taking advantage of smaller producers (Dairy Co, 2011). Although Sanjuan and Dawson (2003) were using different products, there are many aspects of the methodology, which can be used to examine the price transmission of milk in the UK, particularly the use of structural breaks that may occur. By incorporating a break date in to the econometric tests Sanjuan and Dawson (2003) were able to judge whether this period of time (in 1996) had a big affect on the relationship between retailer and producer prices. A weakness of this study is that they could have gone further and examined the ECM, which would allow them to see the speed of recovery back to the equilibrium after the shock had occurred. 20 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 The Common Agricultural Policy In 2003 there was a reform to the European CAP which changed how subsidies were distributed to dairy producers within the UK (Lelyon et al., 2008). This undoubtedly would have implications for the dairy sector as production costs are always increasing and therefore less subsidies would have a massive implication on producers. Zrakic et al (2015) investigated the implications of the 2003 CAP reform on the Croatian dairy industry. By using a simulation model and inputting policy, macroeconomic variables and producers pricing it is possible to forecast the future of the dairy industry. The results found that by 2025 productivity would increase by 25% and the dairy industry in Croatia would be in a more favorable position than before the 2003 reform (Zrakic et al., 2015). It was also suggested that in order to obtain the full benefits of the reform then dairy farmers would have to utilise funds from both the pillar 1 and pillar 2 CAP's. A limitation of using a simulation model like this is that the researchers are only predicting what is going to happen and they cannot allow for any external variables, which could have an effect on the dairy industry, for example Brexit. The data inputted in to the model is based on projections and therefore may be inaccurate and unreliable. Another general limitation of the CAP research is that there are very little studies on the 2003 reform in relation to the milk industry particularly for the UK, thus providing a gap for research. Dairy Retailers Within the UK 40% of raw milk sales are from four main supermarkets, which demonstrates the oligopolistic market (Dairy Co, 2011). A small number of large companies absorb a majority of national milk production. Dairy Co (2011) found that bargaining power, which works in relation to the size of firms, was one of the overriding benefits supermarkets could impose on producers. The main goal for the retailers is to satisfy the consumers, it means they do not proritise with producers (Dairy Co, 2011). The market failure of retailers not transmitting prices downstream to producers in some countries has lead to increased poverty and lower food security (Schroeder and Hayenga, 1987). Retailers offer contracts to producers however Dairy Co (2011) identified weaknesses within these contacts. These include no price certainty, long notice periods and no details on future negotiations. All these factors weaken the position of the producer and it is claimed that milk contracts are simply a "licence to supply" (Dairy Co, 2011). Dairy Producers British dairy farms have been struggling recently and their major concern is that retailers are not paying them a fair price. Farmers have been forced to close their businesses down or even go as far as pouring milk away because they are loosing so much money. Steffen and Spiller (2013) looked in to the efficiency of dairy producers and factors that could be hindering their performance. It was believed that if milk producers were not efficient then 21 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 they would struggle to make appropriate returns. The results found that one of the main factors contributing to lack of efficiency of dairy farmers was their willingness to adapt techniques and unite together to achieve a targeted milk quota for the future. Steffen and Spiller (2013) believed that increased efficiency throughout the supply chain there would enable dairy producers to be more profitable, even with reduced prices. This suggests that the producer is at fault for recent hard times in the milk industry. de Fatima Oliveira et al (2014) originally opposed this view and believed the price paid to producers had a bigger influence on the milk industry than other factors such as efficiency. This was the reasoning behind their research in to the price transmission of milk within the Portuguese market. They found that when the price of the retailer changes the price paid to the producer did not. This suggests that in Portugal it is not the price that is causing hardship on farms and therefore theory that efficiency is to blame for poor milk price return that Steffen and Spiller (2013) proposed, seems feasible. Bor et al (2014) conducted research in a similar manor to de Fatima Oliveira et al (2014) but for Turkey instead of Portugal. The conclusions contrasted as Bor et al (2014) found that large retailers in Turkey act quickly when the input prices of milk increase but they are slower to react when the inputs decrease. This implies that in Turkey large retailers hold all the power shown by the asymmetric price transmission. It also means the retailers control the producer prices and consumer prices, which is the opposite to the Portuguese milk market. The differences between price transmissions across countries are expected because of the individual markets within the country. Bakucs et al (2014) found that policies, governance, laws, economies and power all bare an effect on a countries agricultural markets. This means that the differences across countries, even though evaluating the same product, are normal. Due to these differences occurring horizontally across countries much research starts by looking vertically initially. The Gap The need to look at the long run relationship of the price of milk between the producer and the consumer is more necessary in the UK due to the issues facing many dairy farmers. The 2003 CAP reform changed how subsidies were distributed to milk producers and there is little research in to the effects of this reform on the price of milk. In addition to this the UK has voted to leave the EU so now is an important time for the domestic milk market as the UK will be creating it's own agricultural policies. If British dairy farmers continue to lose out on price then it could have huge consequences on the whole milk industry. Dairy producers are going out of business and they are blaming it on powerful monopoly retailers for driving prices down and therefore it is necessary to assess how true these claims are. A lack of studies within the UK milk market means there is a need to create foundations which will be useful for policy makers, retailers and producers when moving forwards. In addition, there is significantly more research conducted on the milk markets within foreign countries, which gives them a 22 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 competitive advantage over the UK. The varied and mixed results found from other countries in previous literature means there is a big gap for research within the UK. Research indicates there was a change in the milk market in 1994 when the MMB disbanded. There is a need to see if this is highlighted as a structural break and if it had any benefits to either party. Other studies researching the price transmission of milk have not considered breaks, which further confirms the need for the research to be conducted. The collection and availability of 'big data' means an appropriate data range is available which has been a limitation of previous studies. METHODOLOGY Research Title The Price Transmission between producers and retailers within the UK milk market. Research Objectives 1. Establish whether a long run relationship exists between producer and retailer milk prices. 2. Investigate the direction of causality between producer and retailer milk prices. 3. Investigate the effects of structural breaks within producer and retailer milk prices. Data Qualitative data is descriptive data, which can be collected via interviews, surveys or by using secondary data. The data is valuable for giving insight and explanation when conducting research. Many previous studies have successfully used qualitative data to provide great depth and reasoning to endorse their analysis. However qualitative data is not suitable for the econometric tests being used when analysising the price transmission of milk. Limitations of qualitative data are that it is hard to interpret and it is not easy to gather data over a large time scale (Silverman, 2011). These limitations rule out using qualitative data in this research. This research will benefit from using quantitative data, as this is suitable for the econometric tests, which will be used to examine vertical price transmission. Using quantitative data will mean objective results will be obtained. This will give a definitive answer to the research question proposed. In addition it will be possible to acquire a large range of data, which will be important for this research. Quantitative data however, does not give the level of insight and detail which qualitative data does, which is a limitation (Silverman, 2011). 23 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 Primary data is collected first hand, which has the advantage of being tailored and personalised to exactly what is required. Collecting primary data is time consuming, which is a limitation of this research. Primary data is often expensive to collect and usually the results have to be manipulated to make the data usable, which can be time intensive. If the data is collected first hand then the researcher can be sure it is trustworthy or can add variables when collecting the data to make it adhere to features of the methodology. It is difficult to obtain a significant amount of observations in order to collect a suitable amount of data (Saunders, 2011). It will not be possible to collect a large range of time series data and because it is already available primary data is not suitable for this research. The benefits of conducting the analysis with secondary data are that it is easy to obtain a large amount of reliable prices for both the consumer and producer (Saunders, 2011). In addition to this with the time constraints of this research, using secondary is the only viable way to gather the range of data required. The data needed is readily available and is to be sourced from the Office for National Statistics, (2016) which is a trusted and accurate resource for secondary data. The milk prices will be collected for a range of 18 years, which is a sufficient length of time to be able to analyse the long run relationship. Time series data is discrete-time data, which will be used to give monthly increments from 1988 to 2016. The advantages of using time series data is that it allows a comparison of two variables at predetermined time intervals and therefore it is possible to see any correlations (Adams et al., 2014). An alternative would be panel data, which is data, which spans space as well as time. An advantage of panel data is that by combining two dimensions the data has more variation and more degrees of freedom (Saunders, 2011). However for this specific research panel data is harder to obtain and because the only variable we require are the retailer prices and the producer prices, time series data will be used. Testing for Non-stationary data The first step will be to check the data is stationary, which proves whether data is reliable and valid. This will be done by checking the data has unit roots, for both of the variables. The tests for this are the PP (1997) unit root test and the ADF (1979) unit root test. The tests will help establish if there is a trend in the data or whether there are any extreme values. The reason for using two tests is so that we can be absolutely sure the data is valid and reliable. Previous studies have used only one of these tests, which can raise questions over the quality of their data. DeJong et al (1992) argued that the PP test had less power in practice than the ADF test; therefore it is necessary to conduct both. Nonstationary variables mean that there could be statistical issues, like spurious regression or non-sense regression, when analysing a time-series (Cuthbertson and Nitzsche, 2005). This would mean that further statistical interpretation may seem in unity with theory, however the results are not valid and not reliable (Greene, 2012). For this reason it is 24 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 vital to firstly prove stationarity. Stationary time series' can be described as having a constant mean, constant variance and constant autocovariances for each given lag (Brooks, 2014). Using Eviews econometrics software it is possible to conduct both the PP test and ADF tests for both the retailer prices and the producer prices to gain instant results (Griffiths et al., 2012). Firstly the data will be tested in level on Eviews, if the variables are nonstationary then the data will be tested for the first difference (Griffiths et al., 2012). In addition it will be important to run the unit root test with breaks, which are shocks within the data. This will confirm that the data is stationary even with the shocks included, which enhances the reliability. It will only be possible to move on to the cointegration tests once it is proved the data is stationary (Maddala and Lahiri, 2009). Bai-Perron Bai and Perron (2003) investigated structural change models for a range of different elements including the techniques used to select the quantity of breaks, the consistency of break dates and the tests involved in identifying structural changes. The Bai-Perron test can find multiple structural breaks using a bivariate analysis of a relationship. This will give an indication of whether there are structural breaks that occur as a result of the relationship between retailer and producer prices (Bai and Perron, 2003). It will also give up to 5 separate breaks, while other additional tests concerning breaks in the relationships can only show up to two breaks. This is a benefit when using such a large time series of data. Co-integration When it is proved that the variables are stationary then the second test will be to check if the two variables have a long run relationship. The initial analysis of cointegration is to investigate whether it actually exists within the data. The test to examine this is called Johansen (1988) and it tests for long run relationships regardless of breaks or shocks, which may occur within the data (Greene, 2012). Johansen (1988) test starts with the VAR model: Yt = p + A1Yt-1 + A2Yt-2 + ■■■ + ApYt-p + et (1) In equation (1) Yt simultaneously represents both the variables which are integrated in order I(1), producer prices and retailer prices. The VECM is then created: AYt = v + r1AYt-1 + r2AYt-2 + ■■■ + rp-1AYt-p+1 - nYt-± + £t, (2) where: 25 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 rt = -I + A± + A2+-----+ Ai is the matrix for each differenced lag. For 1 = 1,2, ...,k- 1 and n = I - At + A2 + ■■■ + Ak Johansen (1988) uses two key statistics for testing for cointegration, the trace statistic and the maximum eigenvalue statistic. The likelihood ratio tests used to acquire the statistics are: Trace Statistic = -Tln(l - 1) (3) Maximum Eigenvalue Statistic = -Tln(l - ^r+1) (4) For equation (3) the null hypothesis tested is that there are at most r cointegrating vectors present. This means the number of cointegrating vectors is < r, when r equal to 0 or equal to 1. For both values of r the null hypothesis is examined against the general alternative hypothesis. For equation (4) the null hypothesis of r = 0 is tested against the alternative hypothesis of r = 1, then the null hypothesis of r = 1 is tested against the alternative hypothesis of r = 2. If Johansens (1988) test shows one cointegrating vector it means there is a long run relationship between the retailer prices and the producer prices and that one mutual trend is causing the comovement of the two price variables (Chang et al., 2004). The Johansen test can then be conducted with breaks, which ensures that spikes or shocks within the data are not affecting the cointegration. Engle-Granger Cointegration Engle-Granger (1987) is one of the most widely used and reputable cointegration tests (Maddala and Lahiri, 2009). Engle and Granger (1987) stated that after proving both variables (retailer prices and producer prices) are stationary in first levels I(1) we can estimate the cointegration regression by OLS. yt = C + axt + et (5) After identifying the residuals seen in equation (5), the second step is to examine them through a unit root test. This is done by using the PP (2003) test, and if the residuals are stationary then it can be concluded that there is a long run relationship between retailer prices and producer prices. Error Correction Model Cointegration indicates the presence of an ECM. This model establishes how long it takes for the variables to return to a new equilibrium after a shock has occurred (Maddala and Lahiri, 2009). This is used to understand the 26 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 speed of recovery, which will provide further understanding about the relationship between the two variables. Additionally, it will enable suggestions to be made on how to improve future policies. Granger Causality Once the long run relationship is established it is necessary to check the relationship in the short run (up to 5 years), as it may not yield the same results as the long run relationship. The Granger, (1969) causality test will examine the short run relationship and test if there is: • Unidirectional causality - the price of the producer affects the price of the consumer. • Unidirectional causality - the price of the consumer affects the price of the producer. • Bilateral causality - both the price of the producer and the price of the consumer affect each other. • Independence - no relationship between the price of the producer and the price of the retailer. This will establish the direction of causality, which is necessary in this research to understand which variable, retailer or producer price, is having an effect on the other. Grasping this causality will enhance the ability to make future suggestions on the milk market. Momentum Threshold Autoregressive Enders and Granger (1998) and Enders and Siklos (2001) developed the M-TAR model, which tests for asymmetries. The MTAR model is given by equation (6). A|jt = ItPtA^t-i + (1 - /t)p2+ Ipi Yi*Ht-j + £t (6) p1 and p2 are the coefficients which signify the different speeds of adjustment when there is a divergence from the long run relationship (equilibrium). We test for the null hypothesis of no cointegration by using the equation p1=p2=0 in an F-test. The critical values come from Enders and Siklos, (2001). If a cointegration relationship exists then we apply an F-test to p1= p2 with the null hypothesis of symmetry to determine whether asymmetries exist. Empirical Data Figure 1: Natural Logarithms of the Producer and Retailer Prices, UK milk, 1988 - 2016 27 Advances in Business-Related Scientific Research Journal, Volume 9, No. 1, 2018 5.5 3.9 Ln in o o 5 o 59 o o 59 o o 59 o o 59 o o o o CTl