Price Integration of the Ukrainian and EU Corn Markets in the Context of the Russian—Ukrainian War

Abstract

Russia’s full-scale aggression against Ukraine has led to profound disruptions in local and global agri-food markets. Since Ukraine is one of the world’s largest maize exporters, the war also contributed to considerable changes in trade reallocation, as well as an increase in the significance of the European Union in Ukrainian exports. This study analyses the effects of the Russian–Ukrainian war on horizontal maize price transmission between Ukraine and the EU countries. The panel autoregressive distributed lag model (ARDL) was applied to investigate the impact of the war on the price pass-through between those countries. The econometric analysis was performed on a weekly feed maize export price series for Ukraine and 14 selected EU countries. The time frame of research, January 2019 to December 2024, was split into pre-war and war periods. The study indicates that with the outbreak of the war, the long-term relationship between Ukraine and the EU’s maize prices has weakened. At the same time, there was an increase in the short-run maize price transmission between Ukraine and the Eastern EU countries. This proves that in the face of the conflict, market participants in these countries are increasingly guided by the market situation in Ukraine when making economic decisions.

1. Introduction

Ukraine is one of the world’s most dynamically developing producers and exporters of grain, with the dominant maize sector in its production and export structure [1]. In the 2022/23 season, the share of corn in Ukrainian grain exports was 58.8% (27.12 million tons), while wheat was only 37.3% (17.22 million tons). In recent years, Ukraine was the world’s fourth-largest exporter of maize, with an approx. share of 15% (in 2000, this share did not exceed 0.5%). In subsequent years, exports to the EU countries were facilitated by an increase in export quotas at the stage of Ukraine’s accession to the Deep and Comprehensive Free Trade Areas (DCFTA) in 2016. Spain, Italy, and the other EU importing countries could meet a significant part of their demand with cheaper raw materials from Ukraine compared to imports from Brazil or Argentina.

The outbreak of war in Ukraine had several consequences for the functioning of both local and global agri-food markets. The reduction in agricultural exports and increased uncertainty translated into a surge in world agri-food prices and concerns about food security. This problem was widely discussed in the global literature [2,3,4]. The response to Russia’s aggression against Ukraine included a number of initiatives and political actions that led to frequent changes in the direction of grain exports from Ukraine and a significant increase in the export of grains from Ukraine to the EU countries [5,6,7]. The openness of the EU market for the inflow of grains from Ukraine, significant price fluctuations, and market uncertainty led to mass protests by farmers in the EU and to the suspension of imports. It can be assumed that during the war, the nature of the spatial integration of Ukrainian grain markets with world channels has undergone profound changes. It has been manifested in significant trade reallocations [8], with the EU countries, particularly the countries of Eastern Europe, playing a key role.

Spatial market integration determines the efficient allocation of resources and economic welfare. One of its dimensions is price market integration, which, through price transmission studies, reflects how domestic markets respond to supply and demand shocks in other countries [9,10]. Despite the great importance of the efficiency of agricultural commodity markets and their spatial integration for the issue of social welfare, the impact of the war in Ukraine on the transmission of agricultural commodity prices has not been extensively discussed in the literature. Studies conducted by many authors (e.g., [11,12,13]) indicated a strong link between Ukrainian grain prices (mainly wheat) and world prices before the outbreak of the war. However, the strength of price integration between Ukrainian grain markets and world markets has weakened as a result of the war [6,14,15]. In the opinion of those researchers, since 28 February 2022, grain prices in Ukraine have reflected logistic and political risks rather than global supply and demand fundamentals. Hamulczuk et al. [6] found no causality between wheat and maize prices in Ukraine and European and American prices in the first year of the conflict, possibly due to policy interventions and structural changes that weakened the price linkages.

Therefore, the question arises whether and to what extent the war and the accompanying trade creation and trade diversion effects have translated into the strength of the links between maize prices in the EU and respective prices in Ukraine. Our study, in contrast to the above-cited sources, focuses on a broad group of EU countries divided into Eastern EU and Western EU countries. Different price linkages were expected between these two groups. We also excluded from the analysis the very first months of the war period and concentrated on the time since the signing of the Black Sea Grain Initiative (BSGI). This study is also one of the very few that raise the issue of price transmission between the Ukrainian and foreign maize markets, as previous studies have focused solely on the wheat market. Our research fills the gap in this area and contributes to the broadening of knowledge in the field of time-varying price market integration due to global commodity shocks.

2. Implications of War in Ukraine for Agri-Food Markets—A Short Literature Review

The literature on the consequences of the Russian invasion of Ukraine for agri-food markets is relatively extensive. It includes press materials, reports, and scientific studies. In a vast majority of cases, it is devoted to problems of global food security resulting from the blockade of Ukrainian ports in the first phase of the war. As pointed out by Steinbach [8], the first year of the Russian–Ukrainian war resulted in a considerable reduction of Ukrainian exports and led to significant trade reallocation worldwide, which had consequences for food security. Food security issues for net importers have been raised, e.g., by [4,16,17]. In a vast majority of cases, negative consequences were indicated mainly for less developed countries, in particular the Middle East and North Africa (MENA). Al-Saidi [2] drew attention to the varied negative consequences of the conflict for MENA countries. Countries such as Algeria, Turkey, and Jordan coped best, while Egypt and Tunisia were considered the most exposed.

Iacovino et al. [3] emphasised the broad global impact of the armed conflict on the financial situation and food security. This conflict highlighted the existing links between countries, regardless of their geographical location and economic development. For example, Rabbi et al. [18] indicated that the prolonged Russian–Ukrainian conflict would influence domestic food production due to the growing cost of food production, affecting domestic food prices and, consequently, food security across the EU. As an example, they showed the potential adverse effects of the Russian invasion on the production of animal feed and pork in Spain, the largest net importer of feed grains in the EU. Some authors (e.g., [19]) pointed to the negative effects of the Russian invasion of Ukraine on the economic situation in the EU countries. A deterioration of macroeconomic indicators was expected, including a decline in GDP and an increase in unemployment, in the conditions of the current sanctions imposed by OECD countries on Russia and restrictions on importing energy resources. In their research, Kornher et al. [20] referred to the Russian–Ukrainian conflict in terms of its impact on inflation processes in the eurozone countries. As they pointed out, food price fluctuations were primarily the result of the geopolitical situation, which influenced the increase in the costs of energy and food products.

In addition to pointing out the threat to food security, the authors emphasised the importance of global policy aimed at reducing the adverse effects of war on global supply chains, especially for MENA countries. The openness of the EU market (solidarity lines, free trade) played a significant role in cushioning global market shocks and reducing the pressure on the growth of prices of agri-food products [20,21]. One of the most substantial international initiatives aimed at stabilising markets and solving the problem of food security during the conflict was the Black Sea Grain Initiative (BSGI). Its conclusion led primarily to calming sentiment and reducing price volatility. This was a breakthrough, because it was after its signing that grain prices began to gradually decline and grain export returned to pre-war levels [22,23]. Carter and Steinbach [24] confirmed the positive impact of establishing EU solidarity corridors and the signing of BSGI on stabilising grain futures prices on global markets. The announcement of solidarity corridors by the European Union initiated a decline in futures prices to levels before the introduction of these tools. So, in their opinion, the impact of BSGI on prices was secondary. After termination of the BSGI and imposing unilateral bans on imports by neighbouring countries, Ukraine established a new “grain export corridor” hugging the western Black Sea coastline and passing through the territorial waters of Romania and Bulgaria. Besides the new maritime corridor, Ukraine also uses land routes and the Danube River for grain exports, with Romania playing a key role in facilitating these routes. As indicated by Potori and Molnar [7], maize exports exploiting the new grain corridor emerged as the preferred choice, noticeably mitigating market pressures on Eastern EU countries.

Many authors have emphasised the negative consequences of the war for agricultural and food prices. However, the observations and conclusions drawn by the authors depend on the time of the research (conflict phase) or the time perspective of the research. World and local agri-food prices increased sharply at the beginning of the war, surpassing the price levels during the speculative bubble or the so-called Arab Spring [25,26]. Simulations conducted immediately after the outbreak of the war [27] indicated that the export blockade of agri-food commodities in Ukraine and Russia would result in a significant short-term increase in global agri-food prices. However, in the long run, the effects should be much smaller due to global supply responses, which is consistent with previous observations. Reality confirmed the transitory effect of war on agricultural commodity prices. The consequences of the war in Ukraine for world food prices were also studied by Hamulczuk et al. [28]. The most substantial impact of the war on food retail prices was recorded in Europe, Central Asia, and South Asia. The region’s distance from the area of the military conflict, the share of agricultural commodities and energy and fuels in the consumer price, or the conflict response policies pursued by countries were the factors most significantly influencing retail food price reactions.

The war has produced an increased uncertainty in financial markets. The results of analyses based on regression models with heterogeneous effects confirmed the immediate reaction of financial markets to the outbreak of the conflict [29]. According to those authors, the Russian–Ukrainian war led to much greater price volatility in global markets than the global financial crisis and the COVID-19 pandemic. The war also increased financial risk harmonisation between stock returns of G7 countries and commodity prices [30]. Ihle et al. [31] indicated that the war led to disruptions in global supply chains, increasing the synchronisation of grain, energy, and fertiliser prices, creating a global commodity shock. This is confirmed by the studies of Just and Echaust [32] and Capitani and Gaio [33]. The authors emphasised an increase in volatility spillover between energy and grain prices after a military conflict outbreak. However, as emphasised by [24], this was not herd behaviour, and markets put a fair price on the wartime risk of grain shipment disruptions.

Much less research has been done on the spatial linkages of agricultural commodity prices across countries as a result of the war in Ukraine. This is mainly due to the short period, low data frequency (except for futures markets), and high requirements of econometric models. Turvey et al. [14], relying on daily futures data, highlighted significant changes in the differences (the so-called bases) between wheat prices in Ukraine and in the Black Sea region with respect to US prices quoted on the Chicago Mercantile Exchange (CME). At the outbreak of the war and the termination of the BSGI agreement, this basis became negative, which can be attributed to the risk increase caused by the blockade itself and the increase in trade costs. For the same reason, even larger negative differences appeared between grain prices in Ukraine and respective world prices [6]. Turvey et al. [14] confirmed the time-varying Granger causality between these prices, which they attributed to the war and other market conditions.

Cao et al. [15], using daily futures price data from May 2021 to July 2023, investigated how the Russian–Ukrainian war impacted global wheat markets. Using the Bai–Perron test, they identified five market regimes with different price adjustments. They concluded that the linkage of Ukrainian and Russian wheat prices to world prices weakened significantly during the war, especially in the first months. Apart from export restrictions and substantial transaction cost increases, ref. [34] attributed this to a weakened exchange rate pass-through. Nevertheless, Russia’s strong influence on Ukrainian prices after the outbreak of the war continues to be noticeable [15]. Others pointed out that in the first year of the war, not only did the differences between world grain (wheat and maize) prices and Ukrainian domestic grain prices increase, but there was no significant price transmission between them [6]. Thus, the strength of price integration between Ukrainian grain markets and world markets weakened as a result of the conflict. As it follows from these papers, the war led to a temporary disruption of price integration between Ukraine and global grain markets. The above-mentioned authors pointed to the reallocation of price shocks: regional markets partially took over the previously dominant influence of the global market. Grain prices in Ukraine reflected logistic and political risks to a greater extent than global supply and demand fundamentals

3. Materials and Methods

We used a weekly feed maize price series for Ukraine and selected EU countries from January 2019 to December 2024 to investigate horizontal price transmission processes. All data were expressed in euros/tonnes. For Ukraine, we utilised FOB price quotations according to APK-inform. As for the assessments in the EU countries, quotations available on the European Commission (EC) websites were used. They usually concern specific locations and are given in various trade formulas. The EC database contains over 50 time series; however, only a small fraction can be used in time series analyses due to significant data gaps. As a result, price series representing individual EU countries, in which the share of missing data was negligible (up to 5%), were selected. In the case of Germany, the average price was calculated from two locations: Köln and München. The missing values were interpolated based on neighbouring observations. Every data interpolation method is imperfect and introduces unknown biases into the results. Generally, the smaller the number of missing data points, the smaller the impact of interpolation on the results. Therefore, this study uses two approaches, which are averaged: fit piecewise cubic polynomials (AKIMA) and linear interpolation. Additionally, for some countries, we have simultaneous quotations for different locations; thus, it was necessary to choose only the most representative price series for each country. As a result, we used a price series for 14 EU countries from selected locations.

We performed price transmission analyses both for the whole EU panel and for two subgroups: Western EU countries (Belgium (Brussels)—BE, Germany (an average of Köln and München)—DE, France (Rhin)—FR, the Netherlands (Rotterdam)—NL, Portugal (Lisboa)—PT, Spain (Badajoz)—ES, Italy (Bologna)—IT) and Eastern EU countries (Hungary (Great Plain)—HU, Poland (average country price)—PL, Romania (Banat)—RO, Bulgaria (Pleven)—BG, Slovenia (Ljubljana)—SL, Croatia (Zagreb)—CR, Greece (Ioannina)—GR). It was expected that the price transmission may vary depending on the proximity to the Ukrainian border and the maize trade volume between Ukraine and EU countries. It’s worth recalling that the war saw significant changes in the grain trade. Until the outbreak of the war, Ukrainian grain exports to eastern EU countries were practically negligible [6]. With the outbreak of the war and the removal of trade barriers, Ukrainian corn exports to Central and Eastern European countries increased significantly, accounting for approximately 28% of total EU grain imports from Ukraine in 2022–2023. Eastern EU countries also serve or have served as transit routes of goods to other EU or non-EU countries. This could have significantly impacted the corn price linkages. Figure 1 presents Ukrainian and EU average maize price developments for the above-mentioned groups.

To investigate possible heterogeneous price transmission across countries/regions in relatively short time series spans, we applied the panel autoregressive distributed lag model (ARDL) [35]:

$${{∆(EU}_i)}_t={\beta }_0^i+{\sum }_{j=1}^p{{\gamma }_j^i∆\left({EU}_i\right)}_{t-j}+{\sum }_{j=0}^p{{\delta }_j^i∆\left(UA\right)}_{t-j}+{\phi }^i{\left[{(EU}_i\right)}_{t-1}-\{{\beta }_1^i\ast t+{{\beta }_2^i\left(UA\right)}_{t-1}\}]+{ϵ}_{it}$$

where: EU_it—log maize price series in the i-th EU country in time t; UA_t—log maize price series in Ukraine in time t; β—parameters of long-run relationship; φ—speed of adjustment to long-run relationship; γ, δ,—short-run adjustments; j—the number of lags (j = 0, 1, …, p); ϵ_it—random component.

The ARDL model is a very versatile approach because it can be used with variables that are either stationary or non-stationary at the level, it allows for the analysis of both short-run and long-run relationships between variables, and it provides a framework for cointegration analysis. The choice of a panel model over a univariate/multivariate model was dictated by the relatively short research sample—panel models are more powerful in small samples. The applied panel ARDL model also offers various estimators that allow for the adoption of alternative assumptions about the heterogeneity of individual coefficients.

Relying on the performed causality tests (Section 4.2), we assumed stronger causality from Ukrainian (UA) to the EU prices than vice versa. Therefore, the dependent variable was the EU countries’ price series, whereas the independent variable was the Ukrainian price series. The model assumes the existence of a long-term relationship with a constrained trend. The trend is included in long-run relationships to capture changes in trade costs manifested in time-varying price spreads (see Figure 1).

The models were estimated according to three procedures [36,37]. The Pooled Mean Group (PMG) estimator assumes homogeneous long-run relationships across countries, whereas short-run adjustments and error variance are heterogeneous. This estimator might be useful when there are reasons to expect that the long-run equilibrium relationships between variables are similar across countries—that is, in the case of EU grain prices. The Mean Group (MG) estimator estimates different coefficients for each country that are later aggregated. The third estimator, Dynamic Fixed Effects (DFE), assumes that there is no heterogeneity among the countries and thus all coefficients (short- and long-run) are the same for all cross-sections. Only a country-specific intercept is assumed.

In general, the PMG model is more efficient than the MG model, which is related to estimating a smaller number of parameters. This is important in the case of data limitations. On the other hand, the MG model considers the heterogeneity of countries to a greater extent, but is less efficient due to the need to estimate a larger number of parameters. When selecting an estimator, it is worth analysing the theoretical context of the study to predict whether we expect different impacts in individual countries. A decision can also be made based on the Hausman test, which facilitates the identification of the efficiency and consistency of each estimator over others [38]. Basically, it helps decide if the long- and short-run coefficients are homogeneous across all groups or heterogeneous. The Hausman test compares the estimates from a more efficient estimator with those from a less efficient but consistent estimator. If the Hausman test statistic is not significant, it suggests that the long-run homogeneity assumption is valid, and the PMG estimator is preferred.

To investigate the impact of the war on price transmission, panel ARDL models were estimated for two subperiods: before the war (January 2019–February 2022) and during the war period (August 2022–September 2024). Due to the extraordinary situation, the period covering the first 5 months of the war (March–July 2022) was excluded. It was a turbulent time with frequent regime changes and high uncertainty (closing of UA ports, opening of solidarity lines by the EU, or UA currency (hryvna) devaluation). Immediately after the outbreak of the war, seaports were blocked and maize exports from Ukraine practically ceased. Only the launch of the EU–Ukraine Solidarity Lines, the removal of import tariffs and quotas by the EU and the signing of the Black Sea agreement calmed the situation. During this short period, significant structural changes, to which ARDL models are sensitive, could have occurred. As Hamulczuk et al. [6] point out, in the initial period of the war, there could even have been a negative correlation between prices in Ukraine and the EU. Hence, the timeframe for the analysis in this study covers the period from August 2022.

The ARDL panel models were estimated for all the EU countries in the group, as well as specific EU country groups, in order to examine the heterogeneous dynamics of price transmission. All analyses were performed on log data, typical of price series analyses, and facilitated the estimation of price elasticities and Impulse Response Functions (IRFs). The panel ARDL price investigation was preceded by several other procedures based on univariate and panel models.

We started with unit root testing according to the Augmented Dickey–Fuller (ADF) test [39], the panel stationarity of the Im, Pesaran, and Shin test (IPS) [40], and the Breitung test [41]. Knowledge of the degree of integration of the series is important in specifying ARDL models and helps to avoid spurious relationships in empirical estimations. Then we tested Granger causality according to the Toda–Yamamoto (T-Y) procedure [42] and the Dumitrescu and Hurlin test [43]. This makes it possible to confirm the dominant direction of market impulse flow and model specification. In order to verify the existence of a long-run cointegration relationship between UA and EU maize price series, we applied the Westerlund test [44,45]. The panel Westerlund test based on error correction models in small samples tends to exhibit greater statistical power than residual-based panel cointegration tests. We used the bootstrap panel cointegration test statistic, particularly robust in applications involving integrated variables and cross-sectional dependence. All calculations were performed using STATA and Gretl software.

4. Results and Discussion

4.1. Preliminary Analysis

Using log-transformed price series, we started our analysis by testing the unit root for individual countries and different aggregates. In the first step, we used the ADF test to test a unit root in the analysed price series.

Table 1. Unit root testing results using the Augmented Dickey–Fuller test (ADF).

Country	Levels				First Differences
	Model with Constant		Model with Constant and Trend		Model Without Constant		Model with Constant
	01.2019–02.2022	08.2022–12.2024	01.2019–02.2022	08.2022–12.2024	01.2019–02.2022	08.2022–12.2024	01.2019–02.2022	08.2022–12.2024
BE	0.01 [1]	−2.67 [1]	−2.29 [1]	−1.34 [1]	−9.70 [0] ***	−8.85 [0] ***	−9.83 [0] ***	−9.18 [0] ***
BG	−0.13 [2]	−1.79 [1]	−2.5 [3]	−0.84 [1]	−5.95 [4] ***	−9.15 [0] ***	−6.22 [4] ***	−9.31 [0] ***
DE	−0.52 [0]	−2.10 [1]	−2.87 [0]	−1.14 [1]	−13.53 [0] ***	−5.96 [2] ***	−13.58 [0] ***	−15.91 [0] ***
FR	−0.08 [4]	−1.94 [1]	−2.10 [4]	−1.42 [1]	−7.22 [3] ***	−9.79 [0] ***	−7.36 [3] ***	−9.87 [0] ***
GR	0.63 [0]	−1.62 [0]	−1.20 [0]	0.11 [0]	−13.26 [0] ***	−6.11 [1] ***	−13.51 [0] ***	−10.19 [0] ***
HU	0.48 [4]	−1.70 [1]	−2.27 [0]	−1.22 [1]	−13.88 [0] ***	−12.85 [0] ***	−7.93 [3] ***	−12.88 [0] ***
IT	0.94 [4]	−2.21 [1]	−1.68 [4]	−0.98 [1]	−5.97 [3] ***	−5.83 [0] ***	−6.36 [3] ***	−6.07 [0] ***
NL	−0.38 [4]	−2.30 [1]	−3.27 [0]	−0.78 [1]	−8.58 [4] ***	−12.41 [0] ***	−8.80 [4] ***	−12.59 [0] ***
PL	−0.78 [2]	−1.76 [0]	−2.61 [2]	−0.42 [0]	−6.58 [1] ***	−6.08 [1] ***	−6.61 [1] ***	−6.29 [1] ***
PO	1.31 [0]	−1.71 [0]	−1.18 [0]	−0.26 [0]	−8.11 [0] ***	−11.85 [0] ***	−8.38 [0] ***	−12.03 [0] ***
RO	−0.72 [4]	−1.92 [4]	−2.18 [4]	−0.90 [4]	−9.60 [3] ***	−10.82 [2] ***	−9.65 [3] ***	−10.88 [2] ***
SL	−0.16 [3]	−1.42 [1]	−1.87 [3]	−1.29 [1]	−11.30 [2] ***	−13.49 [0] ***	−11.37 [2] ***	−13.46 [0] ***
SP	−0.01 [2]	−1.77 [0]	−1.85 [2]	−1.57 [0]	−6.73 [1] ***	−11.43 [0] ***	−6.97 [1] ***	−11.61 [0] ***
UA	−0.50 [1]	−2.50 [3]	−2.25 [1]	−0.95 [3]	−10.78 [0] ***	−5.33 [2] ***	−10.91 [0] ***	−5.42 [2] ***

Note: *** indicates significance at 1%.

presents the tau test statistic along with the number of autoregressive lags (in brackets). The results indicate that all price time series are integrated of order one. The null hypothesis of a unit root was not rejected for the levels, but was rejected for the first differences of the series. Therefore, we can conclude that all variables are integrated in order 1 (d = 1).

The IPS and the Breitung panel unit root test confirmed these conclusions for the panel series (results upon request). The above conclusions apply to both subperiods, the entire panel, as well as the panels for individual regions. Accounting for the trend did not affect the significance of the test statistic calculated on the series in levels. Due to the cross-sectional dependence confirmed by the Pesaran CD test, the robust Breitung panel unit root test (model with constants) was also used. In the case of Ukraine, the series for all analysed subperiods are integrated in order one (for p = 0.05). The Breitung panel unit root test, robust to cross-sectional correlation, indicates that most EU panels are also integrated in order one. The only exception is Eastern EU prices in the pre-war period, where the null hypothesis of unit roots in the panels was rejected. The non-stationarity was driven by an upward trend observed from June 2020 to February 2022. At the time of Russia’s invasion of Ukraine, prices in the EU countries surged dynamically in early March 2022, but since April 2022, a downward trend has been observed (Figure 1).

4.2. Causality and Cointegration Analysis

In the next step, we investigated whether EU prices are leading Ukrainian maize prices or vice versa (

Table 2. Granger causality according to the Toda–Yamamoto test.

Period	January 2019–February 2022		August 2022–December 2024
EU Country	H0: UA ≠> EU	H0: EU ≠> UA	H0: UA ≠> EU	H0: EU ≠> UA
PL	4.58 [3] **	2.66 [3] *	4.92 [2] **	1.04 [2]
HU	16.08 [2] ***	0.96 [2]	3.03 [3] *	4.21 [3] **
RO	11.48 [2] ***	0.56 [2]	5.78 [4] ***	1.13 [4]
BE	5.02 [2] **	2.24 [2]	1.56 [4]	2.15 [4]
FR	9.83 [2] ***	1.76 [2]	3.02 [3] *	2.82 [3] *
DE	6.42 [2] **	2.11 [2]	1.07 [4]	2.08 [4]
NL	8.78 [2] ***	7.08 [2] **	2.30 [3]	3.00 [3] *
IT	2.97 [3] *	1.46 [3]	0.66 [3]	5.63 [3] **
PO	0.93 [2]	27.42 [2] ***	3.43 [2] *	3.38 [2] *
SP	46.45 [1] ***	0.19 [1]	1.59 [2]	0.48 [2]
BG	5.47 [4] ***	2.30 [4]	1.67 [2]	0.20 [2]
SL	6.14 [3] ***	1.18 [3]	6.83 [2] **	0.76 [2]
GR	4.82 [1] *	0.47 [1]	2.13 [2]	1.27 [2]
Sum of F-stat.	128.97	50.39	37.99	28.15

Note: *, **, and *** indicate significance at 10%, 5% and 1%, respectively.

). It is clear that before the war, the dominant direction of causality came from Ukrainian prices to EU prices, as indicated by the sum of F statistics (128.97 vs. 50.39) of the Toda–Yamamoto test.

The null hypothesis stating that Ukrainian maize prices are not a Granger cause for EU maize prices (H0: UA ≠> EU) was rejected for all cases. Assuming the opposite direction of causality (H0: EU ≠> UA), the null hypothesis was rejected in only three cases. After the outbreak of the war in Ukraine, the direction of causality is not so obvious. However, price signals in most of the analysed pairs of price series still come to a slightly greater extent from Ukrainian to EU prices rather than vice versa. This is evidenced by the higher sum of F statistics for the null hypothesis stating that Ukrainian prices are not a Granger cause for EU prices (37.99 vs. 28.15). In six cases, the null hypothesis about the lack of influence of past changes in EU prices on current changes in EU prices was rejected. An opposite situation (H0: EU ≠> UA) occurred in five cases. In general, it can be concluded that after the war outbreak, Ukraine has lost its importance as a reference point for prices in the EU compared to the pre-war period. Although the invasion slightly weakened the dependence of EU prices on Ukrainian prices, the link with Ukrainian prices has persisted. It is worth mentioning that in the first phase of the war (until mid-2023), refs. [6,15] could not confirm Granger causality between the EU and Ukrainian grain prices. Turvey et al. [14], analysing Ukrainian and world wheat prices, stress that the weakening and changing flows of causality suggest that war resulted in more elastic cross-price trade elasticities between local Ukrainian wheat and the rest of the world.

The durability of the EU price relationship with Ukraine was ensured by the BSGI agreement, while after its completion, by a new export corridor from the ports of Odessa and Pivdennyi (Chornomorsk). In addition, Ukrainian grains were transhipped through the Romanian port of Constanta, which significantly increased the volume of exports [46]. After the outbreak of the conflict, Ukrainian prices had the most substantial causal influence (H0 rejected for p < 0.05) on prices in the countries of Central and Eastern Europe: Romania, Poland, and Slovenia, which served as key transit routes for Ukrainian grain exports in the conditions of logistical bottlenecks and uncertainty of exports through Black Sea ports.

Similar conclusions could be drawn from panel Dumitrescu and Hurlin causality tests (

Table 3. Granger causality according to the Dumitrescu and Hurlin test.

Specification		January 2019–February 2022		August 2022–December 2024
EU Group	Statistic	H0: UA ≠> EU	H0: EU ≠> UA	H0: UA ≠> EU	H0: EU ≠> UA
EU	W-bar	16.742	5.193	9.063	4.192
	Z-bar	27.580 ***	11.094	21.332 ***	4.101
	Z-bar tilde	26.813 ***	3.902	20.678 ***	3.902
Western EU	W-bar	17.100	9.347	3.810	5.873
	Z-bar	19.975 ***	15.616 **	2.394	5.123
	Z-bar tilde	19.439 ***	15.254 **	2.270	4.908
Eastern EU	W-bar	20.334	1.039	14.155	2.511
	Z-bar	36.171 ***	0.073	24.612 ***	0.676
	Z-bar tilde	35.362 ***	0.049	23.876 ***	0.610

Note: ** and *** indicate significance at 10%, 5% and 1%, respectively.

). In order to obtain more robust results, the causality tests were performed using the bootstrapping method, enabling the consideration of heterogeneity. The number of lags (1 or 2) was selected according to BIC, whereas p-values were computed using 400 bootstrap replications. Before the war, UA prices were a Granger cause for EU and Eastern EU prices. In the case of Western EU countries, the direction of causality was two-sided, with a slightly greater impact of UA on EU prices rather than vice versa. During the war, the null hypothesis that UA maize prices are not a Granger cause for EU maize prices was rejected for the EU and the Eastern EU aggregates. The test indicates a lack of any Granger causality between prices in Ukraine and Western EU countries (which does not exclude the occurrence of a long-run relationship and mutual price determination).

In the next stage of the study, the existence of a long-term relationship between the Ukrainian and EU maize price series was tested using the Westerlund panel procedure.

Table 4. Testing panel cointegration using the Westerlund test.

Period	January 2019–February 2022				August 2022–December 2024
Statistics	Gt	Ga	Pt	Pa	Gt	Ga	Pt	Pa
EU
Stat	−3.275	−21.510	−13.799	−24.961	−2.799	−12.725	−10.711	−13.174
Z-val	−6.238	−9.875	−8.432	−17.472	−4.253	−3.837	−5.325	−7.538
p-val	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
p-val (robust)	0.000	0.000	0.000	0.000	0.000	0.010	0.000	0.000
Eastern EU
Stat	−3.182	−22.105	−9.936	−26.194	−3.260	−16.749	−8.329	−15.740
Z-val	−4.136	−7.272	−6.141	−13.089	−4.366	−4.669	−4.525	−6.859
p-val	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
p-val (robust)	0.000	0.000	0.000	0.000	0.000	0.000	0.003	0.008
Western EU
Stat	−3.453	−21.827	−9.032	−21.332	−2.338	−8.701	−5.870	−8.261
Z-val	−4.933	−7.137	−5.232	−10.192	−1.649	−0.757	−2.052	−2.402
p-val	0.000	0.000	0.000	0.000	0.050	0.224	0.020	0.008
p-val (robust)	0.000	0.000	0.000	0.000	0.085	0.213	0.070	0.099

presents the testing results with 1 lag (selected according to AIC) and p-values computed using 400 bootstrap replications. The table shows four statistics (Gt—group-t statistic, Ga—group-a statistic, Pt—panel-t statistic, Pa—panel-a statistic). In the case of Gt and Ga, the null hypothesis assumes no cointegration for at least one cross-sectional unit. In turn, in Pt and Pa, we tested the null hypotheses stating a lack of cointegration in the entire panel. For the EU aggregate, the null hypothesis is rejected in four test statistics, both before and during the war. Similar results are found for the Eastern EU panel. However, for the Western EU, the results are less clear. While the null hypothesis for the pre-war period, assuming the lack of cointegration, is rejected, the conclusion is less evident for the war period. For three statistics, the p-value suggests rejection of H0 at the significance level of p = 0.05. The robust p-value indicates rejection of H0 for these test statistics (Gt, Pt, Pa) only at the level of p = 0.10. In the case of Ga, there is no basis for rejecting the null hypothesis stating no cointegration for at least one cross-sectional unit.

4.3. Price Transmission Based on Panel ARDL Results

Results based on the ARDL panel models are of key importance for our analyses. The estimated coefficients for the panel ARDL (1.1) models are presented in

Table 5. ARDL model estimation results.

Specification	PMG	MG	DFE	PMG	MG	DFE
	January 2019–February 2022			August 2022–December 2024
EU
EUt−1	0.848 ***	0.914 ***	0.897 ***	0.678 ***	0.721 ***	0.739 ***
trend	0.001 ***	0.001 ***	0.001 **	−0.002 ***	−0.002 ***	−0.002 ***
EC	−0.153 ***	−0.166 ***	−0.159 ***	−0.137 ***	−0.173 ***	−0.139 ***
ΔEUt−1	0.009	0.007	−0.252 ***	−0.013	0.001	−0.247 ***
ΔUAt	0.179 ***	0.188 ***	0.187 ***	0.263 ***	0.301 ***	0.334 ***
ΔUAt−1-	0.005	−0.012	0.035	−0.020	−0.035	0.015
const	0.112 ***	0.050	0.081	0.296 ***	0.274 ***	0.264 ***
Eastern EU
EUt−1	1.066 ***	1.020 ***	0.980 ***	0.933 ***	0.843 ***	0.847 ***
trend	0.000	0.001 *	0.000	−0.001 ***	−0.002 ***	−0.002 ***
EC	−0.153 ***	−0.177 ***	−0.165 ***	−0.178 **	−0.208 **	−0.167 ***
ΔEUt−1	−0.107 *	−0.099 *	−0.290 ***	−0.053	−0.043	−0.272 ***
ΔUAt	0.111 *	0.122 *	0.094	0.323 ***	0.344 ***	0.356 ***
ΔUAt−1-	−0.024	−0.038	−0.009	−0.085 *	−0.090 *	−0.031
const	−0.079 **	−0.089	−0.004	0.115 **	0.170	0.212 **
Western EU
EUt−1	0.813 ***	0.809 ***	0.798 ***	0.610 ***	0.599 ***	0.608 ***
trend	0.001 ***	0.001 ***	0.001 ***	−0.002 ***	−0.002 ***	−0.002 ***
EC	−0.153 ***	−0.155 ***	−0.134 ***	−0.124 ***	−0.139 ***	−0.104 ***
ΔEUt−1	0.122 *	0.112 *	0.057 *	0.037	0.045	−0.077 **
ΔUAt	0.257 ***	0.254 ***	0.244 ***	0.254 ***	0.258 ***	0.281 ***
ΔUAt−1-	0.012	0.015	0.024 **	0.024	0.021	0.027
const	0.160 ***	0.188 **	0.149 ***	0.328 ***	0.379 ***	0.276 ***

Note: *, **, and *** indicate significance at 10%, 5% and at 1%, respectively.

. We applied unified lags for all estimators and subperiods. The models were estimated with a lag of 1, while alternative specifications with lags of up to four periods were tested. Ultimately, the most conservative specification with a first-order lag was adopted because further expansion of the number of lags did not significantly improve the test statistics, including the information criteria (AIC, BIC), and all additional lagged coefficients were insignificant.

Most of the estimated coefficients are statistically significant. All the coefficients for the EU_t-1 variable describing long-term relationships have reasonable values and are statistically significant at p < 0.001. The trend included in the long-term relations also turned out to be significant in most cases (except for Eastern EU before the war). The coefficients controlling the speed of adjustment to the long-run relationship (EC) have the expected signs and are statistically significant. Therefore, the estimated coefficients can generally be considered relatively reliable and intuitive. The number of lags used was also sufficient to eliminate the autocorrelation of the residuals. According to the literature [35], correctly selecting the number of lags in an ARDL model is considered an effective method for correcting both autocorrelation and endogeneity problems. This practice is widely accepted in studies using panel ARDL models, confirming the dominant role of the model’s dynamic structure in ensuring estimation accuracy.

The three alternative estimators give fairly consistent results for key statistics, which indicate the robustness of the research results. Namely, there are no large differences in the estimates of long-run relationships (EU_t−1) or price adjustments to such relationships for the same periods (EC). For example, during the war, these coefficients for the EU aggregate range from 0.678 to 0.739 and from −0.173 to −0.137, respectively. Additionally, slightly faster price adjustments to the long-run relationship can be observed for the MG estimator rather than the PMG and FE estimators. The Hausman test [47] used by the authors to determine the choice between PMG, MG and DFE estimators provided ambiguous results. In most cases, there were no statistically significant differences between these estimators, both before and during the war. Given the economic context of the study and the fact that, due to the lack of trade barriers, grain markets in EU countries are highly integrated, it is expected that in relatively homogeneous groups (eastern and western parts of the EU), a similar long-run relationship between prices in Ukraine and prices in EU countries may also occur. This suggests that the PMG estimator is a worthwhile option. Pesaran et al. [37] also drew attention to the better efficiency of the PMG model in the case of homogeneous countries in terms of economic development.

Based on the estimated models, the cumulative impulse response functions (IRFs) were calculated and presented in

Table 6. Cumulated IRFs based on estimated panel ARDL models.

Period	Before the War			War Period
Region	EU	Eastern EU	Western EU	EU	Eastern EU	Western EU
Horizon (weeks)	PMG model
0	0.18	0.11	0.26	0.26	0.32	0.25
1	0.29	0.22	0.39	0.30	0.33	0.33
4	0.51	0.52	0.58	0.43	0.59	0.43
8	0.67	0.76	0.71	0.54	0.77	0.50
12	0.76	0.90	0.77	0.60	0.85	0.55
16	0.80	0.97	0.79	0.64	0.90	0.58
20	0.83	1.01	0.80	0.65	0.92	0.59
24	0.84	1.04	0.81	0.66	0.92	0.60
Horizon (weeks)	MG model
0	0.19	0.12	0.25	0.30	0.34	0.26
1	0.30	0.23	0.38	0.34	0.34	0.34
4	0.56	0.55	0.58	0.51	0.59	0.44
8	0.74	0.79	0.70	0.62	0.74	0.51
12	0.83	0.90	0.76	0.67	0.80	0.55
16	0.88	0.96	0.79	0.70	0.82	0.57
20	0.90	0.99	0.80	0.71	0.84	0.59
24	0.91	1.01	0.80	0.72	0.84	0.59
Horizon (weeks)	DFE model
0	0.19	0.09	0.24	0.33	0.36	0.28
1	0.29	0.20	0.36	0.32	0.31	0.32
4	0.48	0.46	0.52	0.45	0.51	0.40
8	0.65	0.67	0.65	0.56	0.65	0.47
12	0.75	0.80	0.72	0.63	0.73	0.51
16	0.81	0.87	0.75	0.67	0.78	0.54
20	0.85	0.92	0.77	0.69	0.81	0.57
24	0.87	0.94	0.79	0.71	0.83	0.58

. In general, it can be seen that the strength of price transmission between Ukraine and the EU aggregate starting from August 2022 has been weaker than in the three years before the war, regardless of the estimator used.

According to the panel ARDL models, a 1% shock in maize prices in Ukraine during the war resulted in a 0.43–0.51% change in the EU after 4 weeks and 0.54–0.62% after 8 weeks. Before the war, these responses were 0.48–0.56% and 0.65–74%, respectively. In the long run, the EU price responses to the shock in UA prices during the war were 0.16–0.19 pp. weaker than before the war. This is consistent with the results of [15], who also noted that despite the resumption of maritime grain exports, the strength of the linkages between Ukrainian and world wheat prices is still lower than in the pre-war period. It should be emphasised that the conflict has significantly reduced Ukraine’s agricultural production potential, and consequently its export capacity [48]. At the same time, major global grain exporters, including Russia, have increased production, pushing Ukraine out of some markets. In other words, weakening Ukraine’s export potential and quick adjustment of the world markets into a new spatial equilibrium have translated into international maize price linkages.

The panel model estimation in groups of countries gives a more diversified picture. Firstly, regardless of the analysed sub-period (pre-war or war periods), in the long run, prices in the Eastern EU countries are more strongly linked to Ukrainian prices than it is in the case of the Western EU (Table 5 and Table 6 and Figure 2). Secondly, during the war (since August 2022), there was a weakening of the price transmission for both regions in the long run. However, the transmission strength decreased more in the Western EU (e.g., for the PMG model from 0.813 to 0.610) than in the Eastern EU (from 1.066 to 9.333, respectively). Thirdly, short-term price adjustments have changed. They are manifested both in the form of the coefficient standing in front of the EC term (Table 4), and in the graphical form in Figure 2. Namely, during the war, price adjustments to the long-term relationship in the Eastern EU became stronger than before the war, while in the Western EU, they weakened. As a result, currently, both short-term and long-term price reactions to changes in Ukrainian prices are faster in the Eastern EU compared to the Western EU. For example, during the war, a 1% change in Ukrainian prices after 4 weeks resulted in a 0.59% change in prices in the Eastern EU and a 0.43% change in prices in the Western EU. Before the war, these reactions were 0.52% and 0.58%, respectively. It seems that such an increase in short-term price sensitivity in the CEECs is due to the large inflow of maize from Ukraine to these countries compared to the pre-war period, when maize imports from Ukraine were practically non-existent. The growing importance of Eastern European countries in transporting and retaining Ukrainian grain was also indicated by [49]. It is worth emphasising that the increase in price transmission strength is taking place even though some Central and Eastern European countries introduced an import ban on grains from Ukraine in the spring of 2023, which is why direct imports in some countries have almost disappeared.

The cumulative impulse response functions (Figure 2) illustrate the increase in short-term price responsiveness in the Eastern EU to Ukrainian export prices and the decrease in long-term dependence between them. These results align with intuition, especially in the context of the growing role of those countries in grain trade and thanks to their close location [7,49]. Despite the subsequent import ban, the inflow of maize from Ukraine to neighbouring countries caused grain markets in the eastern EU to react relatively quickly and nervously to the events in Ukraine. The increased speed of price transmission means that farmers and other market participants in these countries are increasingly guided by the market situation in Ukraine when making economic decisions than before the war. This also signals that the ongoing institutional integration between Ukraine and the EU may result in even stronger trade and price linkages.

5. Conclusions

This study aimed to assess the effects of the Russian–Ukrainian war on horizontal price transmission processes between the Ukrainian and EU maize markets. The study was conducted using the panel ARDL model, which facilitated the formulation of several conclusions. First, the study confirmed a long-run (cointegration) relationship between maize prices in the EU and UA both before and during the war. However, the null hypothesis of a lack of cointegration for the Western EU countries was rejected at a lower significance level for the Eastern EU aggregate. At the same time, the coefficients illustrating long-run relationships in the Eastern EU countries were higher than those for the Western EU countries. This can be associated with the longer distance and greater dependence of Western European countries on maritime transport, which was relatively strongly affected by the blockade of ports and uncertain international agreements (BSGI). Secondly, the study showed a weakening of long-term relationships between UA and the EU prices after the outbreak of the war (as observed both for the western and eastern EU countries). This was probably influenced by frequent changes in market regimes and various market interventions. Thirdly, the direction of Granger causality between prices in UA and prices for the EU aggregate and the Eastern EU countries did not change due to the war. That is because the dominant direction of market impulse flow remained from Ukrainian prices to EU prices. However, in the case of the Western European countries, the null hypothesis of no Granger causality was not rejected (in any direction). Fourthly, the effect of the outbreak of the war was connected with an increase in the short-term maize price transmission between Ukraine and the Eastern EU countries. Countries close to Ukraine began to react faster to market impulses in Ukraine, which was not the case before the war. The proximity of the Ukrainian market, the increase in imports in the first dozen or so months of the war, the use of these countries for re-export, and farmers’ protests increased the responsiveness of maize prices in these countries to price changes in Ukraine.

In general, this study provided more insight into how war impacts horizontal maize market integration during the shock caused by war. On the one hand, the outbreak of the war led to an increase in trade with the EU countries, which can be considered a manifestation of the growth of spatial market integration. On the other hand, this did not clearly translate into an increase in price links between Ukraine and the EU countries. An exception is found for the growth in the short-term price transmission between Ukraine and the Eastern EU countries, and only in the short term. Market participants in the Eastern EU countries increasingly consider prices in Ukraine when making decisions. The nature of competition between Ukraine and Eastern European countries has changed. Until the outbreak of the war, competition between them in grain markets had been indirect through global trade links. During the war, direct links also emerged, as Ukraine was forced to export grain to or through these countries. It demonstrates that the war led to a reorganisation of the maize trade in Ukraine and the European Union, thus leading to changes in price relationships between them. Observing the agricultural consequences during the war also allows us to draw some conclusions about the ex-ante effects of the deepening processes of Ukraine’s integration with the European Union. It can be assumed that further trade liberalisation or Ukraine’s accession to the EU will permanently change the geographical scope of grain markets, leading to stronger trade and price links, especially with neighbouring countries.

Key limitations of this study include a relatively small data sample and low frequency of data being utilised, which did not allow us to investigate price transmission processes in shorter regimes during the war. Weekly frequency data, compared to daily data, do not allow for a full examination of potential structural breaks over a sample period of less than three years. Due to frequent changes in policy interventions during wartime, time-varying price transmission can also be expected. Another issue worth examining is the potential heterogeneity of price transmission, which can be investigated using univariate or multivariate time series models. Equally interesting could be research on non-linear price adjustments between Ukraine and European Union countries. Non-linear modelling can address both long-term relationships and short-term price adjustments.