448
Views
6
CrossRef citations to date
0
Altmetric
Research Article

Price level convergence in Turkey

ORCID Icon & ORCID Icon

ABSTRACT

This paper examines price convergence among 26 regions of Turkey using general and disaggregated monthly consumer prices between 2005 and 2019. In contrast to previous studies, the results provide evidence for convergence clubs (multiple equilibria) and support the heterogeneity of consumer prices across Turkish regions. This finding is particularly valid for the disaggregated prices where there exist two to four clubs. There is also an increase in the number of convergence clubs in recent years. That is, market integration or the law of one price in Turkey does not appear to be achieved not only for the prices of nontradable goods but also for those of tradable goods either, most likely due to a combination of factors including higher exchange rate volatility, rising inflation and unrelenting regional socio-economic disparities. Our results show that the inflation-targeting regime does not help converge the consumer price levels across the regions.

JEL CLASSIFICATION:

I. Introduction

This paper examines price convergence at the aggregate and disaggregated level across 26 regions of Turkey by employing a workhorse approach for convergence testing, namely the log t test developed by Phillips and Sul (Citation2007), Phillips and Sul (Citation2009); hereafter, PS). There is a good reason to expect the prices of identical goods converging in the long-run because the law of one price or strong market integration generally holds in a single market, that is, price differences of homogenous products or factors in spatially separated markets do not exceed transaction costs such as transportation costs; otherwise, arbitrage opportunities arise. It is also argued that monetary policy regime can have a strong impact on price convergence. Turkey is an exciting case because of two characteristics. After a long history of high inflation, Turkey has started to implement an explicit inflation-targeting regime in 2006, which has brought inflation rates down to single-digit levels in the following ten years. Hence, it is valuable to understand the extent to which the inflation-targeting regime has contributed to the price level convergence across regions in Turkey (Yilmazkuday Citation2013). It is equally vital to understand the regional disparities in price levels if the central bank wants to avoid asymmetric or unintended consequences and welfare loss. Second, it is a member of the G-20 with large socio-economic and territorial imbalances (see ). Thus, inflation rates across its regions can reflect such heterogeneity. Differences in inflation rates can also persist due to the Balassa-Samuelson effect (i.e. heterogeneities in the relative productivity growth of the tradable vs the nontradable sector) or different degrees of market integration or climate, among others.

Figure 1. The map of Turkey: regional socio-economic development disparities.

This map presents the results of a composite index for monitoring regional socio-economic development levels at NUTS 2 level. Light blue and darker blue regions respectively represent higher and lower development scores. Regional codes include following cities: R10(Istanbul); R21(Tekirdag, Edirne, Kırklareli); R22(Balıkesir, Canakkale); R31(Izmir); R32(Aydin, Denizli, Mugla); R33(Manisa, Afyonkarahisar, Kutahya, Usak); R41(Bursa, Eskisehir, Bilecik); R42(Kocaeli, Sakarya, Duzce, Bolu, Yalova); R51(Ankara); R52(Konya, Karaman); R61(Antalya, Isparta, Burdur); R62(Adana, Mersin); R63 (Hatay, Kahramanmaras, Osmaniye); R71(Kırıkkale, Aksaray, Nigde, Nevsehir, Kirsehir); R72 (Kayseri, Sivas, Yozgat); R81(Zonguldak, Karabuk, Bartin); R82(Kastamonu, Cankirı, Sinop); R83(Samsun, Tokat, Corum, Amasya); R90 (Trabzon, Ordu, Giresun, Rize, Artvin, Gumushane); RA1(Erzurum, Erzincan, Bayburt); RA2(Agrı, Kars, Igdır, Ardahan); RB1(Malatya, Elazig, Bingol, Tunceli); RB2(Van, Mus, Bitlis, Hakkari); RC1(Gaziantep, Adiyaman, Kilis); RC2(Sanliurfa, Diyarbakır) and RC3(Mardin, Batman, Sirnak, Siirt). Source: https://www.bebka.org.tr/?lang=en
Figure 1. The map of Turkey: regional socio-economic development disparities.

Our work is closest to Yilmazkuday (Citation2013). However, the current study extends it in the following two aspects. Our data span from 2005 to 2019, while Yilmazkuday’s research covers the period between 1994 and 2004. That is, we cover the whole period of the explicit inflation-targeting regime.Footnote1 His dataset includes ten expenditure-based consumer price indices (CPI) for seven geographical regions (NUTS1 level), while ours have 12 CPI sub-groups for 26 provinces (NUTS 2 level). Second, in contrast to the pair-wise approach in Yilmazkuday (Citation2013), this paper applies the PS log t test to test for (relative) convergence and then use their clustering algorithm to reveal whether convergence clubs are available.Footnote2 Phillips and Sul (Citation2007) also employ CPI to illustrate the usefulness of their method for the convergence in the cost of living across US metropolitan areas. Consequently, this study will be one of the first empirical studies employing PS methodology to analyse price level convergence in emerging markets such as Turkey.

II. Data and empirical analysis

Data used to measure regional prices are monthly aggregate and disaggregated CPI across 26 regions of Turkey from 2005 to 2019 and obtained from the Turkish Statistical Institute. displays the statistical regions together with regional socio-economic disparities in Turkey. The disaggregated level data cover 12 expenditure-based CPI groups for each region: (1) Food and non-alcoholic beverages (2) Alcoholic beverages and tobacco, (3) Clothing and footwear, (4) Housing, (5) Furnishings and household equipment, (6) Health, (7) Transportation, (8) Communication, (9) Recreation and culture, (10) Education, (11) Hotels, cafes and restaurants, and (12) Miscellaneous goods and services. All indices are based on 2003.

Before we apply the log t test, we transformed all series into logarithms. Then, we isolated the cyclical component and extracted the trend component of consumer prices using the Hodrick-Prescott filter.gFootnote3 To avoid the base year problem, we also discarded some fraction of the sample suggested by PS.Footnote4

We present our empirical results of the log t test in . Panel A of the table provides an estimate of the slope coefficient equal to −0.63 and t-statistic of −21.72, which strongly indicates the rejection of absolute price convergence across 26 regions at the 1% significance level. Furthermore, the clustering algorithm identifies two clubs and two divergent regions in the case of aggregate consumer prices. The first club (Club 1) consists of eighteen regions with an estimated beta coefficient of −0.313 and an estimated t-value of −15.686, while the second club (Club 2) includes six regions with an estimated beta coefficient of 0.584 and the estimated t-value of 57.605. Two diverging regions include Antalya, Isparta, Burdur; and Mardin, Batman, Sirnak, Siirt.Footnote5 These results do not change when we further investigate the probability of merging clubs to avoid the possibility of over-estimating the number of clubs. It is, however, noteworthy that more than 80% of the total urban population are clustered in Club 1, which broadly suggest that the law of one price holds for most people and areas across the country. Nevertheless, there seem to be two different clubs of regions with varying levels of convergence speed.Footnote6

Table 1. The log t test results for general consumer prices across Turkish regions.

presents the initial and final classification of convergence clubs for consumer prices at the disaggregated level. Only health prices have a single club.Footnote7 Prices of housing, water, electricity, gas and other fuels, and alcoholic beverages and tobacco have two clubs, but most of the regions are members of only one club. These results are not surprising because the government regulates these markets. Nevertheless, heterogeneity is a common characteristic of the rest of the disaggregated consumer prices. The prices of main expenditure groups classified as nontradable goods and services such as education, recreation and culture, and hotels, cafes and restaurants have three convergence clubs and a few divergent regions. Other services components, including transportation and communication, have three clubs as well. Furnishings and household equipment have two clubs. More interestingly, tradable goods such as food and non-alcoholic beverages, clothing and footwear are grouped into three clusters. The above findings do not bode well with those of Yilmazkuday (Citation2013), who finds that while the CPI groups with tradable components diverge from each other, the CPI groups with nontradable items converge under the inflation-targeting regime.

Table 2. Summary of convergence clubs by main categories of the consumer prices.

We further investigate the evolution of convergence clubs through an expanding window estimation of the PS procedure. It indicates how the pattern of convergence changes over time. We only report the year-end results of the analysis in , as it is sufficient to present the overall picture. We cannot reject the null hypothesis of full panel convergence for the general consumer price index before 2016. Sub-convergence of the general consumer prices appears noticeable in the last two years, albeit most of the regions are still in the same club. The results for the most disaggregated consumer prices, on the other hand, suggest a varying number of convergence clubs over time and hence reject the absolute price convergence for most of the periods. Food and non-alcoholic beverages are an exception. It starts with a single group similar to the general CPI but then includes two clubs after 2013 and ends up in three clubs in 2018 and 2019.

Table 3. Recursive club convergence analysis.

We conjecture that the Turkish Lira’s massive volatility and depreciation played an essential role in explaining our recent findings (see, among others, Ozdemir (Citation2020). Its depreciation rate over the last five years has exceeded thirty per cent in real terms. Due to the high pass-through effects of the exchange rates on consumer prices, Turkey’s inflation reached double digits in 2017 and even exceeded 20% in 2018. Therefore, as Yilmazkuday (Citation2013) argued, recent regional price level differences most likely reflected the high volatility in the exchange rates depending on the degree of exchange rate pass-through specific to the region.Footnote8 Higher inflation figures and higher exchange rate volatility in recent years seemed to contribute to the divergence of consumer prices across regions, with the most negligible impact on the general price level. Besides, the widening positive gap between actual and targeted inflation led to the loss of credibility of Turkey’s central bank, built up in the initial periods of inflation targeting regime.

III. Concluding remarks

This paper investigated the price convergence across 26 regions in Turkey using monthly general and disaggregated consumer prices between 2005 and 2019. In contrast to most previous studies employing panel unit-root tests or traditional convergence methodologies, we employed the log t test. We found the rejection of absolute convergence of consumer prices and the existence of club convergence, both at the aggregate and the disaggregated level. We observed two convergence clubs at the general price level, albeit one of which primarily covers most regions. There are mainly two to four clubs at the disaggregated price levels depending on the CPI component. These findings refer to lower market integration and are in contrast to previous studies. Moreover, lower market integration, across tradable and nontradable goods, partly reflects Turkey’s continued regional disparities. For instance, the regions with a lower level of development, such as those in the southeastern part of Turkey, generally are members of the same club.

Furthermore, a recursive analysis of the convergence clubs suggested that the convergence of general consumer prices across Turkish regions in the earlier periods of inflation targeting regime was generally valid for most of the time, a finding foreseen by the law of one price and previously evidenced by Yilmazkuday (Citation2013). However, recent years have witnessed increasing non-convergent behaviour at the general CPI level and the disaggregated levels.

In addition to the Balassa and Samuelson effect, it seems that the higher exchange rate volatility and rising inflation in recent years have played a role in the divergence of consumer prices across the regions. This is contrary to the initial years of the inflation-targeting regime during which inflation rates were relatively under control and exchange rates were much more stable. There is hence a risk that existing price-level divergence may deepen the regional disparities and misallocate productive resources.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Notes

1 Before 2006, the CBRT had adopted an implicit inflation-targeting regime between 2002 and 2005.

2 We consider regional price convergence within the framework of the club convergence hypothesis. Accordingly, if regions have different initial conditions and some other attributes such as initial income levels, the stocks of physical and human capital (e.g. level of literacy), or location (geography, climate), a group of regions may approach a particular equilibrium due to those common drivers and form clusters or clubs. In contrast to previous approaches that select the clubs according to some a priori criteria such as geographical regions, the PS methodology uses endogenized grouping to identify club convergence by not specifying the factors that contribute to multiple equilibria (clubs).

3 We also used the Hamilton filter as a robustness check. The results showed an increase in the number of convergence clubs in a few cases. As the results did not change qualitatively, we proceeded with the HP filter to allow for comparisons with similar studies in the literature. Also, note that the Hamilton filter cuts the first 35 observations and leaves 3,952 observations for the analysis.

4 We discard the initial 30% of the sample. PS suggest that 0.3 is sufficient. The results did not change when we discarded the initial 20% of the sample.

5 Antalya, which is one of two divergent regions, is known to be the most developed province in greenhouse cultivation. The second region (Mardin, Batman, Sirnak and Siirt), has the highest unemployment rates in Turkey. These two characteristics can largely explain their relatively lower consumer price levels and hence their divergence status.

6 The conventional use of east/west dualism in Turkey as a reflection of regional income disparities does not seem sensible here. There is no distinct geographic pattern in the club convergence results of general consumer prices.

7 Turkey has a comprehensive public healthcare system across the country. Moreover, residents can also get medical care at discounted fees from contracted private hospitals as well.

8 Moreover, Turkish Lira’s depreciation could also motivate agricultural producers to export, leading to divergence even in food prices across regions.

References

Reprints and Corporate Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

To request a reprint or corporate permissions for this article, please click on the relevant link below:

Academic Permissions

Please note: Selecting permissions does not provide access to the full text of the article, please see our help page How do I view content?

Obtain permissions instantly via Rightslink by clicking on the button below:

If you are unable to obtain permissions via Rightslink, please complete and submit this Permissions form. For more information, please visit our Permissions help page.