The Analysis of the Life Expectancy and the Selected Causes of Deaths in Poland with the Use of Spatial Statistics Methods

Mielecka-Kubien, Zofia Article The Analysis of the Life Expectancy and the Selected Causes of Deaths in Poland with the Use of Spatial Statistics Methods Comparative Economic Research. Central and Eastern Europe Provided in Cooperation with: Institute of Economics, University of Łódź Suggested Citation: Mielecka-Kubien, Zofia (2012) : The Analysis of the Life Expectancy and the Selected Causes of Deaths in Poland with the Use of Spatial Statistics Methods, Comparative Economic Research. Central and Eastern Europe, ISSN 2082-6737, Łodz University Press, Łodz, Vol. 15, Iss. 4, pp. 161-175, https://doi.org/10.2478/v10103-012-0033-7 This Version is available at: https://hdl.handle.net/10419/259125 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0 10.2478/v10103-012-0033-7 ZOFIA MIELECKA-KUBIEŃ∗ ∗∗ ∗ The Analysis of the Life Expectancy and the Selected Causes of Deaths in Poland with the Use of Spatial Statistics Methods Abstract The goal of the presented research was to test the spatial autocorrelation of the life expectancy and the age-standardized mortality rates for selected causes of death in Poland according to gender in 2010. It was assumed that in the above mentioned cases the positive spatial autocorrelation in populations of men and women appears, and the spatial diversity of mortality depends on the standard of living of the population in question and on the level of industrialization of the region and its consequences. It has been stated that most of the considered coefficients show positive spatial autocorrelation, but differences between populations of men and women were observed. Agricultural capacity of the voivodeship shows positive effect on life expectancy and the level of some of the mortality rates of both genders. 1. Introduction With the development of the methods of spatial statistics, as well as of the geographic information system (GIS), it became possible to conduct the deepened research on spatial variation of mortality characteristics. The aim of the presented study was to verify the supposition – using the basic spatial statistics methods - whether life expectancy (e0) and mortality rates from selected causes of deaths in Poland show spatial autocorrelation. The especially ∗Ph.D., Professor at the University of Economics in Katowice 162 Zofia Mielecka-Kubień interesting question was, whether the pattern of spatial autocorrelation was the same for men and women (living in the same country, at the same time). Finally, an attempt was made to find factors possibly influencing the observed spatial differences and spatial autocorrelation patterns of the coefficients taken into account. 2. The applied method The research was conducted with the use of the basic spatial statistics methods, i.e.: 1. Moran’s global statistic I, defined as follows (Kopczewska 2006, p.72): ( ) ( ) ( ) ∑ ∑ ∑ ∑∑ − −− ⋅= ii i j jiij i j ij yy yyyyw w n I 2 (1) where: n – denotes the number of spatial units indexed by i and j (here voivodeships or in the case of life expectancy, also subregions of Poland), y – the considered coefficient, wij – element of the matrix of spatial weights W, constructed according to numbers of the first order neighbours, rowstandardized. Expected value of statistic I is: ( ) 1 1 − − == n IE µ (2) Assuming that the sample comes from the independent random variables normally distributed, test statistic I s : ( ) ( ) ID IEI I S − = (3) is asymptotic normal distributed 1 ( ) 1,0N ≈ . On this basis the hypothesis Ho : I = 0 against the hypothesis 0I:H 1 ≠ is tested. 1 See for instance (Ekonometria przestrzenna 2010, p.109). The Analysis of the Life Expectancy… 163 Local Moran statistic I i is defined as: ( ) ( ) ( ) ∑ ∑ − −− = ii ijiji i nyy yywyy I/ 2 (4) with the expected value (Anselin 1995) given by: ( ) ∑ = − − = n jiji w n IE 1 1 1 . (5) The local Moran test, based on the conditional randomisation or permutation (see: Anselin 1995), detects the local spatial autocorrelation. There can be two interpretations of the local Moran statistics: as indicators of the local spatial clusters (regions where adjacent areas have similar values) and as a diagnosis for the outliers in global spatial patterns (areas distinct from their neighbours). The Local Moran statistic I i decomposes Moran's I into contributions for each location: ∑ = i i n I I (6) Additionally the Moran scatterplots and maps were applied. Moran scatterplot allows (Anselin 2005) to explore the global patterns of autocorrelation in space. The graph depicts the standardized variable (here life expectancy or age-standardized mortality rates) in the x-axis versus the spatial lag of that standardized variable, where the spatial lag shows the effects of the neighboring spatial units. Moran scatterplot presents the relation of the variable in the location i with respect to the values of that variable in the neighboring locations. By construction, the slope of the line in the scatter plot is equivalent to the Moran's I statistic. If that slope is positive it means that there is the positive spatial autocorrelation: high values of the variable in location i tend to be clustered with high values of the same variable in locations that are neighbors of i, and vice versa. If the slope in the scatter plot is negative it means that high values in a variable in location i tend to be co-located with lower values in the neighboring locations. 164 Zofia Mielecka-Kubień 3. Empirical results The presented research was conducted 2 on the basis of 16 voivodeships of Poland (spatial units NUTS2); in the case of the life expectancy for men and women smaller units (66 subregions, NUTS3) could be taken into account. The data for the year 2010 come from the Chief Statistical Office in Warsaw. All the considered mortality rates were standardized with regard to age. The following variables were subjects of the study for men (m) and women (k): life expectancy in voivodeships (Y om ,Y ok ) and subregions (X om ,X ok ), general mortality rates (Y 1m ,Y 1k ), cancer mortality rates (Y 2m ,Y 2k ), circulatory system diseases mortality rates (Y 3m ,Y 3k ) and respiratory system diseases mortality rates (Y 4m ,Y 4k ). While circulatory system diseases and cancer were the most frequent causes of death in 2010 in Poland (tab.1), respiratory system diseases were chosen with regard to their specific spatial differences pattern. Table 2 presents the results of the testing of the hypothesis about the absence of the spatial autocorrelation on the basis of Moran global statistic I (as described above). It can be observed that not all of the considered variables show significant spatial autocorrelation. Table 1. Percent of deaths for chosen causes in Poland, 2010 DISEASE PERCENT OF DEATHS Circulatory system 46.0 Cancer 25.4 Respiratory system 5.1 Source: author’s own. As can be observed (fig.1, tab.2) the life expectancy for men in voivodeships does not show spatial autocorrelation – different values of life expectancy are randomly distributed across the country. The shortest life expectancy for men in Poland in the year 2010 was observed in the voivodeship Lodzkie; on the contrary, the longest life expectancy for men was in the voivodeships Malopolskie and Podkarpackie, in the southern part of the country. 2 For calculation there were used computer programs R and EXCEL, for visualisation – programs EXCEL and Statistica. The Analysis of the Life Expectancy… 165 Table 2. Results of testing hypothesis of absence of spatial autocorrelation COEFFICIENT MORAN I STATISTIC TEST STATISTIC I S MEN WOMEN MEN WOMEN Life expectancy -0.1191 0.2771 -0.6163 4.0433 Life expectancy, subregions 0.4002 0.6095 4.5198 6.7966 General mortality rates -0.1311 0.1908 -0.7577 3.0285 Cancer mortality rates 0.4194 0.4602 5.7161 6.1962 Circulatory system diseases mortality rates -0.0790 0.1044 -0.1453 2.0123 Respiratory system diseases mortality rates 0.1901 0.2502 3.0201 3.7268 Remark: cases of rejected null hypothesis ( α = 0.05) are marked in bold. Source: author’s own. Figure 1. Moran plot for life expectancy (e 0 ), men Source: author’s own. However, the theoretical elimination of the three outliers (voivodeships: Lodzkie, Malopolskie, Podkarpackie, marked on fig.1 with black triangles) indicates, that apart from the three voivodeships, the life expectancy for men is characterized by negative spatial autocorrelation (y1*om, dashed line), which means, that in the case of the remaining voivodeships it is revealed that the neighboring values are more dissimilar than expected by chance. A considerably different pattern emerges from the men life expectancy considerations based on the subregions of Poland (tab.2, fig.2) – here the spatial autocorrelation is more positive than expected at random, which indicates the 166 Zofia Mielecka-Kubień clustering of similar values across smaller items in geographic space. The longest life expectancy is observed (tab.3) in the big cities: Warsaw, Cracow, subregion trojmiejski containing the cities of Gdansk, Sopot and Gdynia and in the south-eastern corner of Poland (subregions: rzeszowski and tarnowski). The lowest values of men life expectancy can be seen (as in fig.2) in the subregions of the voivodeship Lodzkie, and subregion stargardzki in the northern part of Poland. Figure 2. Spatial differences in men life expectancy (in years), subregions 69 - 71 71 - 73 73 - 74 74 - 76 Source: author’s own. For women both spatial patterns of life expectancy (the one based on the voivodeships as well as the one based on the subregions), are different than those for men (fig.3 and 4). Apparently the Polish women living, generally speaking, in the eastern part of the country enjoy longer life than the ones living in the westren part. In this case both spatial differences patterns are similar, however the more detailed analysis (subregions) uncovered some significant exceptions: Warsaw, Wroclaw and subregion trojmiejski, but, as indicated in table 3, the longest women life expectancy is observed in the eastern subregions of Poland (and not in the big cities as in the case of men). The shortest women life expectancy was observed in the subregions belonging to the voivodeships: Lodzkie and Slaskie. A strong positive spatial autocorrelation can be observed for the men and women mortality of cancer (tab.3, fig.5 and 6). Apparently the inhabitants of the north-western part of Poland are more at risk from cancer, and, in the case of women, especially those living in the voivodeship kujawsko-pomorskie. The lowest values, for both genders, can be observed in the eastern part of Poland – with generally much higher level of age-standardized cancer mortality rates in The Analysis of the Life Expectancy… 167 the population of men. For the age-standardized mortality rates for the diseases of the respiratory system the spatial differences pattern as well as the level of positive spatial correlation (tab.3) are similar for both genders, but the voivodeship Warminsko-mazurskie takes the strongly exceptional position (fig.7) – the mortality rates are very high (especially for men). The theoretical elimination of that outlier could change the slope of the regression line in the Moran scatter plot for men from a 1 = 0.1901 to a 2 = 0.6484 (fig.8, y1*4m, dashed line), and for women from a 1 = 0.2502 to a 2 = 0.6762. Table 3. Subregions of highest and lowest life expectancy (in years) according to gender HIGHEST LIFE EXPECTANCY, MEN HIGHEST LIFE EXPECTANCY, WOMEN 75.3 Warsaw 82.1 bialostocki 75.1 Cracow 82.0 tarnobrzeski 74.6 trojmiejski 81.8 łomzynski 74.4 rzeszowski 81.7 rzeszowski 74.2 tarnowski 81.7 suwalski LOWEST LIFE EXPECTANCY, MEN LOWEST LIFE EXPECTANCY, WOMEN 70.0 lodzki 79.1 grudziadzki 70.0 Lodz 79.1 sosnowiecki 70.0 skierniewicki 79.0 Lodz 70.0 stargardzki 78.8 lodzki 69.9 piotrkowski 78.5 katowicki Source: author’s own. Figure 3. Spatial differences in women life expectancy (in years), voivodships 79 - 80 80 - 81 81 - 81 81 - 82 Dolnośląskie Kujawsko-pomorskie Lubelskie Lubuskie Łódzkie Małopolskie Mazowieckie Opolskie Podkarpackie Podlaskie Pomorskie Śląskie Świętokrzyskie Warmińsko-mazurskie Wielkopolskie Zachodniopomorskie Figure 4. Spatial differences in women life expectancy (in years), subregions 78 - 79 79 - 80 80 - 81 81 - 83 Source: author’s own. 168 Zofia Mielecka-Kubień Figure 5. Spatial differences in cancer mortality rates*, men, voivodships 363 - 387 339 - 363 315 - 339 291 - 315 Dolnośląskie Kujawsko-pomorskie Lubelskie Lubuskie Łódzkie Małopolskie Mazowieckie Opolskie Podkarpackie Podlaskie Pomorskie Śląskie Świętokrzyskie Warmińsko-mazurskie Wielkopolskie Zachodniopomorskie Figure 6. Spatial differences in cancer mortality rates*, women, voivodships 142 - 160 160 - 178 178 - 196 196 - 214 Dolnośląskie Kujawsko-pomorskie Lubelskie Lubuskie Łódzkie Małopolskie Mazowieckie Opolskie Podkarpackie Podlaskie Pomorskie Śląskie Świętokrzyskie Warmińsko-mazurskie Wielkopolskie Zachodniopomorskie Note: * – per 100 000 population Source: author’s own. Figure 7. Spatial differences in respiratory diseases mortality rates*, men 49 - 79 79 - 109 109 - 138 138 - 169 Dolnośląskie Kujawsko-pomorskie Lubelskie Lubuskie Łódzkie Małopolskie Mazowieckie Opolskie Podkarpackie Podlaskie Pomorskie Śląskie Świętokrzyskie Warmińsko-mazurskie Wielkopolskie Zachodniopomorskie * – per 100 000 population Source: author’s own. The Analysis of the Life Expectancy … 175 Streszczenie BADANIE DŁUGOŚCI ŻYCIA ORAZ WYBRANYCH PRZYCZYN ZGONÓW W POLSCE Z ZASTOSOWANIEM METOD STATYSTYKI PRZESTRZENNEJ Celem prezentowanego badania było testowanie hipotezy o braku autokorelacji przestrzennej w odniesieniu do przeciętnego dalszego trwania życia oraz standaryzowanych ze względu na wiek współczynników zgonów dla wybranych przyczyn zgonów w Polsce według płci w 2010 r. Przypuszczano, że w wyżej wymienionych przypadkach występuje dodatnia autokorelacja przestrzenna w populacjach mężczyzn i kobiet oraz, że przestrzenne zróżnicowanie umieralności zależy od poziomu życia danej populacji i stopnia industrializacji regionu i jej konsekwencji. Okazało się, że większość z rozważanych współczynników wykazuje dodatnią autokorelację przestrzenną; zaobserwowano też różnice między populacjami mężczyzn i kobiet. Rolniczy charakter województwa wykazuje pozytywne oddziaływanie na przeciętne dalsze trwanie życia i wartości niektórych współczynników zgonów dla obu płci.