The objective of this study was to analyse the impact of positive externalities of international tourism demand on increasing the market power (MP) of an extra virgin olive oil (EVOO) wholesaler in Croatia. In the context of this article, the MP measures how close the wholesaler can set the actual price of EVOO to the maximum the retailer wants to pay. Our hypothesis explained how the additional demand of tourist consumers for EVOO could stimulate and increase the MP of the wholesalers. Here, it was important to remember that the EVOO market signals relatively asymmetric quality information about products that varies in certain ranges. The selected time-series span the weekly period from 2017 to 2019. We used the Toda-Yamamoto approaches of causality in the relationship between the EVOO price gap and tourism overnights, as well as the autoregressive distributed lag model (ARDL) bounds test for cointegration. For larger EVOO bottles (0,75 and 1 l), there is unidirectional causality flowing from tourism consumption, which we presume originates from the tourism demand variable, to MP. There is a relevant bidirectional causality in the case of the 0,25 l bottle. Tourism in a purchased bottle of 0,5 l does not manifest any side-effect impact on MP. This pioneering study has investigated the relationship between the MP of EVOO wholesalers in Croatia and tourist demand. An inventive view has been adopted with regard to the theoretical concept of measuring MP, but also due to the steps towards the use of ARDL bound testing.

This article investigates the MP of EVOO, which in our opinion is associated with rising international tourist demand, at least until the Covid-19 pandemic, using the example of the Republic of Croatia’s wholesale trade. We argue that purchasing power is enriched by tourist buyers, which in turn causes EVOO price rises, empowering wholesalers’ position in the supply chain. International tourist demand has recorded high growth rates in recent years, which has increased the consumer potential for buying bottled EVOO in Croatia. Using the positive externalities, our hypothesis asserts that tourist consumers cause an increased MP for wholesalers.
The research on market power (MP) in the olive oil trade is relatively scarce and it is mostly centred on exports

The subject of MP in the olive oil trade has not been studied extensively, at least
according to academic articles composed in the past. A comparative analysis of the
MP of olive oil exporting countries indicates that exports are imperfectly
competitive in the EU market, and that Italy has higher MP compared to Spain and
Greece

THEORY

The literature suggests a considerable range of MP measures based on industry
concentration measures, entry barriers, and various indexes (for more one can see

MP = f (+ NIGHT) (1)

Where:

MP is as we suggested, streamlining, ground on the differential between maximal price and minimal price; that gap is a proxy for the MP and NIGHT represents international overnight stays, which on the other hand is a proxy for international tourism demand. Such a formulation is in line with the aforementioned general consideration regarding the link between overnight tourist stays and the MP in the oil wholesaler’s market consideration. Accordingly, the main hypothesis of this study is as follows: an increase in overnight stays involving international tourists boosts the MP of EVOO from a wholesaler perspective. ECONOMETRIC METHODOLOGY

ARDL cointegration and bounds tests

To analyse the long-term relationship between a set of variables, authors

Δ𝑌𝑡=𝑐𝑜𝑛𝑠𝑡+Σ𝛽Δ𝑌𝑡−1𝑘𝑖=1+Σ𝛾Δ𝑋𝑡−1𝑘𝑖=1+𝜔𝑌𝑡−1+ 𝜃𝑋𝑡−1+ 𝜀𝑡 (2)

where Yt denotes price gap measured in EVOO and Xt denotes a tourist overnight stay as a single input (as explained previously), both expressed in natural logarithms. An appropriate lag selection is based on the Schwarz Bayesian Criterion (SBC). The automated model selection process involves choosing the maximum lag for each regressor, which is set as 8 (because the data is weekly). The ARDL procedure allows for the possibility that the variables may have different optimal lags (after the searching process has ended), whereas this is impossible with conventional cointegration procedures. The null hypothesis for no long-term relationship between the price gap and the tourism input variable is not rejected, by testing the 𝐹-statistic, when:

𝐻0: 𝜔=𝜃=0, against the alternative 𝐻0: 𝜔≠𝜃≠0.

Instead of the conventional critical values, authors

Causality analysis

The initiation of proceedings within Toda-Yamamoto causality analysis occurs if the cointegration link between MP and NIGHT persists. Formally speaking, if a price gap and an international tourist overnight input as a metric of consumption are regressed against one another in levels, the resulting residuals essentially represent error correction terms. These terms measure deviations in the long-run equilibrium between the two series. Hence, the ARDL Eq. (1) can be re-parameterised after replacing Yt−1 and Xt−1 with the lagged residuals:

e.g., the error correction model via the two-step procedure of Engle and Granger.

These lagged residuals represent an error correction term, denoted in this article by
ECT (−1), which provides an insight into the speed of adjustment to a long-run
equilibrium within a particular time frame from a change to one of the series.
Furthermore, if the coefficient of ECT (–1) is statistically significant (by
t-value), long-run causality is confirmed. ECTt-1 should be between 0 and 1 with a
negative sign, which implies convergence of the system back to the long-run
equilibrium position. Additionally, the direction of cause and effect between the
variables – testing the hypothesis that tourist demand measured by the number of
overnight stays causes an increase in the MP of EVOO wholesalers – will be clarified
by using the Modified Wald test (MWALD) recycled according to the Toda-Yamamoto
(1995) procedure. The MWALD test skips obstacles and problems associated with the
classic Granger causality test resulting from non-stationarity or cointegration
between series. It is to be expected that the latter problems would cause a
theoretical inconsistency in collision with the empirical performance of the
classical Granger causality test

From Eq. (3), the Granger causality from NIGHTt to MPt implies 0∀i; similarly, in Eq.
(4), MPt Granger-causes NIGHTt if 0∀i. The bi-variable model is estimated using
seemingly unrelated regression (SUR), as in

ABOUT THE DATA

In order to examine the relationship among the variables, the study used weekly
time-series data from 2017-2019 for four different volumes of bottled EVOO. In this
study, the MP proxied by price gap is the dependent variable. Tourism demand is
proxied by overnight stays (NIGHT), and this is the only variable singled out as the
independent variable. MP in our reconsidered theory of price differential
variability provides the basis for its dynamic measurement. The MP marked with
different volume labels – detailed later in the text – will distinguish various
prices per volume: 0,25 litres, 0,5 litres, 0,75 litres, and 1 litre. Auxiliary
variables (maximal and minimal prices) for calculation of MP were sourced from an
overview of wholesale EVOO prices by week given in Kn per litre, given in

UNIT ROOT TEST

Before conducting tests for cointegration, it is vital to ensure that the variables
under consideration have not been integrated at an order higher than 1. In the
presence of I(2) or higher variables, the computed statistics provided by authors

Levels | MP1 | –11,800 (0)*** | –4,022 (1)** |

First diff. | – | – | |

Levels | MP075 | –7,298 (1)*** | –7,185 (1)*** |

First diff. | – | – | |

Levels | MP05 | –3,642 (3)*** | –4,413 (2)*** |

First diff. | – | – | |

Levels | MP025 | –5,480 (2)*** | –7,068 (1)*** |

First diff. | – | – | |

Levels | NIGHT | –2,511 (4) | –2,912 (1)* |

First diff. | –4,988 (4)*** | –5,021 (3)*** |

*Firstly, all the regressions include a linear trend in the levels and include an intercept in the first differences; secondly, the numbers in parentheses are the optimal lag orders and are selected based on Schwarz Bayesian; thirdly, *, ** and *** denote the 10 %, 5 % and 1 % level of significance, respectively.

The DF–GLS test statistics include an intercept and a linear time trend in the levels
and only an intercept in first difference

RESULTS OF THE ARDL COINTEGRATION TESTS

In the first step in applying the bounds test, we specified the optimal lag length
of the UECM, i.e., Equation (1), and checked the long-run level equilibrium
relationship. We have attempted to optimally set up the ARDL model and fixed an
optimal lag, which is crucial. With an initial lag of 8, the aputomated model
selection, according to minimal SBC

Dep. | Indep. | Bounds | Bounds t-test | Cointegration | LM-test | JB-test | HET |

variable | variables | F-test | |||||

statistic | |||||||

MP025 | Night | 40,168 (0,000) | –10,967 (0,000) | Yes | 0,285 | 0,548 | 0,855 |

MP05 | Night | 6,951 (0,001) | –4,485 (0,001) | Yes | 0,031 | 0,643 | 0,62 |

MP075 | Night | 38,444 (0,001) | –10,725 (0,000) | Yes | 0,989 | 0,496 | 0,879 |

MP1 | Night | 39,433 | –10,849 (0,000) | Yes | 0,941 | 0,924 | 0,937 |

0 |

*The critical values for the 𝐹-statistic are derived from table CI (III) (see Table 4).The range for the associated t-test is: lower-bound I(0) = –3,43, upper-bound I(1) = –3,82. LM is the Lagrange multiplier test for serial correlation with a 𝑥2 distribution, with only one degree of freedom; J-B is the Jarque-Bera test for normality; HET is the White test for heteroscedasticity with a 𝑥2 distribution with only one degree of freedom; asterisks *, ** and *** denote statistical significance at the 1 %, 5 % and 10 % levels, respectively.

I(0) | I(1) | |

10 % critical value | 3,17 | 4,14 |

5 % critical value | 3,79 | 4,85 |

2,5 % critical value | 4,41 | 5,52 |

1 % critical value | 5,15 | 6,36 |

The F-statistics for cointegration analysis based on the selected ARDL models are reported in Table 2 for all EVOO price gap cases. All the reported F-statistics – as well as the t-statistics – lie above the upper bounds; consequently, the null hypothesis of no cointegration is rejected and the precondition for cointegration is established in all four EVOO volumes.

SHORT-RUN ESTIMATES

Table 4 indicates the short-run implications of tourism overnight stays on the price gap, e.g. MP of wholesalers. While the EVOO price gap is found to have a lagged impact on itself in the case of 0,25 l (one lag) and 0,5 l (two lags), whereas in the case of the other two bottle profiles (0,75 and 1 l) there is an instantaneous autoregressive short-run impact. The price gap (or MP) affects its own trajectory negatively at a statistically significant level in two cases (in a bottle: volume 0,5 l and 1 l). Overnight stays have a positive and statistically significant lagged impact on price gap, e.g. the MP of wholesalers, in all bottle cases except for the 0,75 l case where that impact is enhanced by the instantaneous die out. One time-lagged error correction term is negative and statistically significant at a 1 % level in the case of all the analysed price gaps for the various bottles.

Volume | Variables | Lags | |||

0 | 1 | 2 | 3 | ||

0,25 l | ΔMP | – | –0,497 | – | – |

(–6,899) | |||||

ΔNOC | 0,001 | 0,097 | 0,023* | 0,507** | |

-0,125 | -0,846 | -1,869 | -2,086 | ||

ECMt_1 | –0,811*** | – | – | – | |

(–7,448) | |||||

0,5 l | ΔMP | –0,372 *** | –0,173** | ||

(–3,967) | (–2,312) | ||||

ΔNOC | 0,376 | 0,451 | 0,047** | 0,866* | |

-0,126 | -0,024 | -2,231 | -1,935 | ||

ECMt_1 | –0,443*** | – | – | – | |

(–4,599) | |||||

0,75 l | ΔMP | 0,89 | – | – | – |

-1,315 | |||||

ΔNOC | 0,590** | 0,887** | 0,342 | –0,414 | |

-2,153 | -2,332 | -1,276 | (–0,743) | ||

ECMt_1 | –0,890*** | – | – | – | |

(–10,823) | |||||

1 l | ΔMP | –0,920 | – | – | – |

(–10,850) | |||||

ΔNOC | 0,181 | 0,455 | 0,955 | 0,144*** | |

-0,845 | -0,873 | -0,821 | -2,561 | ||

ECMt_1 | –0,870*** | – | – | – | |

(–10,945) |

*Asterisks *, ** and *** denote statistical significance at the 1 %, 5 % and 10 % levels, respectively.

LONG-RUN ELASTICITY

We consider weekly measurement of the included variables with 156 observations to be a sufficiently reliable basis to estimate the long-run interference of tourism overnight stays on the price gap in EVOO trading. The long-run elasticity of a single independent variable (NIGHT) with respect to the dependent variable (MP) is shown in Table 4. All EVOO price gaps referring to various sold bottles have statistically significant and positive relationships, and are affected by an increase in international tourism overnight stays. Here, as predicted in the theoretical overview, we can state that tourism demand causes a price gap/MP increase for wholesalers.

NOC |
0,76** (2,311) |
0,756** (2,231) |
0,329*** (2,897) |
0,412* (2,119) |

intercept |
2,447 (1,452) |
1,296 (1,397) |
1,185 (1,264) |
1,826 (1,003) |

*Ibidem.

TODA-YAMAMOTO CAUSALITY TEST

After estimating long-run results, we proceeded to the causality test. However, first the long-run causality is conducted by the coefficient t-statistics, which stand before the ETCt-1, wherein this term measures how fast the deviations from the long-run equilibrium die out following changes in the NIGHT variable. The lagged ECT coefficients from Table 5(4) show that international overnight stays and MP in all types of contracted bottled EVOO were restored to the long-run equilibrium. The analysis of Equation (6), as presented in Table 7, indicates that there is long-run cointegration among the variables at the 1 % significance level; moreover, the ECT coefficients revealed in Table 8 imply that any deviation from the long-run equilibrium is corrected within the adjustment speed range of 44-89 % (the quickest speed is measured for the 0,75 l volume EVOO bottle). This also indicates a strong causality for the tourist demand variable on the MP of EVOO wholesalers. From the estimation of the Toda-Yamamoto Granger causality test (see Table 5), we can make the following assertion based on the results of this study: For 0,75 l and 1 l bottled EVOO there is unidirectional causality flowing from international overnight stays demand to the price gap (MP), thereby supporting our hypothesis that tourism affects MP in EVOO trading. For the 0,25 volume bottle, the reverse causality evidence, where tourism overnight stays is caused by price gap, revealed that the MP also has side-effects on tourism overnights, thus forming a bidirectional causality linkage.

0,25 l | 1 | 1+2 | 5,654 | 0,098* | NIGHT → MP |

1 | 1+2 | 13,455 | 0,018** | MP → NIGHT | |

0,5 l | 1 | 1+2 | 2,756 | 0,446 | NIGHT do not cause MP |

1 | 1+2 | 9,966 | 0,813 | MP do not cause NIGHT | |

0,75 l | 1 | 1+2 | 12,734 | 0,013** | NIGHT → MP |

1 | 1+2 | 0,814 | 0,556 | MP do not cause NIGHT | |

1 l | 1 | 1+2 | 9,388 | 0,024** | NIGHT → MP |

1 | 1+2 | 0,896 | 0,403 | MP do not cause NIGHT |

*The (k+dmax) denotes VAR order. The lag length selection was based on AIC: Akaike information criterion. * and ** denote the 1 % and 5 % significance level, respectively. → Denotes one‑way causality.

This article was directed towards attaining a full understanding of the causal relationships between international tourism overnight stays and the MP of EVOO from the wholesaler’s perspective. For the bottled EVOO products that pass the rigorous statistical testing – the unit root, cointegration, and bounds testing

This article is a result of scientific project Tourism development and destination impacts supported by the Faculty of Economics and Tourism “Dr. Mijo Mirković”, Juraj Dobrila University of Pula. Any opinions, findings, and conclusions or recommendations expressed in this article are those of the author(s) and do not necessarily reflect the views of the Faculty of Economics and Tourism “Dr. Mijo Mirković” Pula.