The relevance of supply shocks for inflation: the spanish case

For Peer Review The Relevance of Supply Shocks for Inflation: The Spanish Case Journal: Applied Economics Manuscript ID: APE-06-0390.R1 Journal Selection: Applied Economics JEL Code: E31 - Price Level|Inflation|Deflation < E3 - Prices, Business Fluctuations, and Cycles < E - Macroeconomics and Monetary Economics Keywords: Inflation, Nominal rigidities, Supply shocks, Skewness, Spanish Regions Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript For Peer Review 1 The Relevance of Supply Shocks for Inflation: The Spanish Case 1. Introduction The idea of the present contribution is based on several factors: i) Spain is a country characterised by a persistent moderate inflation differential with the core EU countries –see European Central Bank (2003). ii) The figures of Spanish inflation have slightly increased in recent years and the Spanish Government faces problems to control inflation –see Bank of Spain (2006). iii) The irregular evolution of oil prices in recent years, with several adverse supply shocks, deserves a lot of international attention –see Kilian (2005). Our paper tries to shed some light jointly on these factors, from the Spanish perspective, proposing some explanations mainly based on the use of Ball and Mankiw’s (1994, 1995) approach. In order to implement panel data techniques and provide additional information at a regional level, we pay special attention to the Spanish regional inflation data, although we also include in our analysis several control variables. Empirical evidence shows that inflation and the higher moments of the distribution of relative prices are positive correlated, against the theoretical predictions of the flexible price model. Ball and Mankiw (1994,1995) show that inflation is mainly influenced by skewness, arguing that, in presence of nominal rigidities, due to the fact that firms face menu costs, changes in the price level and skewness are positively correlated; effect that can be magnified by the standard deviation of the distribution, denoted as relative price variability (RPV) in this strand of the literature. Our study tries to check if the skewness-inflation relation holds for Spain and if the behaviour of Spanish regions is homogeneous with respect to it. The analysis of such relation can be relevant in the sense that these authors show that skewness is a proxy for supply shocks, and therefore that relation is explaining how sensitive the inflation is when a supply shock affects the economy and if a supply shock affects to the same extent all regions. Positive inflation-skewness and inflation-RPV relations are supported by the data, but results are not conclusive about which relation is stronger. On one hand, for Page 1 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 2 periods with an annual inflation rate lower than 4%-5%, the inflation-skewness relation is stronger than the inflation-RPV one –see Ball and Mankiw (1995) for the US, Lourenco and Gruen (1995) for Australia, Amano and Macklem (1997) for Canada, Aucremanne et al. (2002) for Belgium and Caraballo and Usabiaga (2004a,b) for Spain, among others. Moreover, for some high inflation countries there is evidence of a positive association between inflation and skewness, as Raftai (2004) shows for Hungary in a period with an annual inflation rate ranging from 15% to 30%. On the other hand, for studies covering periods with changing inflation rate, the evidence is mixed. For example, Hall and Yates (1998), for the 1975-1996 period in the United Kingdom, find a weaker inflation-skewness relation than the inflation-RPV one for the whole period. More precisely, both relations are stronger for the high inflation period and the former is even negative for the low inflation period, in contrast to the results obtained by Assarsson (2004) and Caraballo and Dabús (2005). The first author finds for Sweden that RPV and skewness are more important in explaining inflation in the low inflation period than in the high inflation one. Caraballo and Dabús (2005) find the same results of Assarsson (2004) for skewness but not for RPV. These authors focus on Spain and Argentina, concluding that RPV is significant for both low and high inflation periods for Argentina and only for the high inflation period in Spain. In addition, they find that skewness is significant for the low inflation period but not in the high inflation period in both countries, even though the mean inflation rate in each period differs strongly across them. In fact, the mean annual inflation rate of Argentina in the low inflation period (around 23%), is higher than the Spanish inflation rate in the high inflation period (14%). Finally, Döpke and Pierdzioch (2003), for the 1969-2000 period in Germany, find that both relations are positive, but none of them is clearly stronger. Table 1 tries to summarise the main empirical evidence on this topic. [Table 1] This mixed evidence can be due to different reasons, and specially to the fact that the relation between inflation and the higher moments of the distribution of changes in relative prices is very sensitive to changes in the features of mean inflation. Page 2 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 3 Generally in low inflation countries both variables are significant but depending on the trend of inflation a relation can be more significant than the other.1 The main contributions of this paper, in comparison with previous ones in this area for the Spanish economy –Caraballo and Usabiaga (1994a, 1994b), Caraballo and Dabús (2005)–, are the following: we work with a higher degree of disaggregation in the data, a very important feature in this kind of literature based on price changes distribution functions; we extend and update the period of analysis; and we incorporate as control variables the main economic variables related to this topic available for the Spanish economy with a monthly frequency. The rest of the paper is organised as follows. Section 2 presents the main data and variables. In section 3 we develop a preliminary analysis for the 17 Spanish regions. Section 4 performs a panel data analysis. In section 5 several control variables are included, and section 6 concludes. 2. Main data and variables Our analysis refers to the 1993.02-2005.12 period. We are aware of the shortness of this sample period (13 years) in comparison with other studies, but it is not possible to extend it, due to the important data requirements of our analytical methodology, with a high degree of disaggregation in the data, as well as the use of several control variables. Only the period considered fulfils all this data matching. However, we have to take into account that the data are monthly, a frequency which is not commonly used in the literature, and consequently we get 155 observations of each series. Our sample period can be clearly divided into two subperiods. The first one goes from 1993.02 to 1998.12, and is characterised by a negative trend inflation, and a mean monthly inflation rate around 0.28%. The second one is the 1999.01-2005.12 period, in which no trend inflation is found and presents a mean monthly inflation rate around 0.26%. 1 See Caraballo and Dabús (2005) for further details. Page 3 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 4 The main data used are the series of monthly change rates of consumer price indexes, disaggregated by goods and services (57 categories), for the 17 Spanish regions elaborated by the Instituto Nacional de Estadística (INE). The weight of each subgroup offered by the INE is defined as the proportion of expense made on that article in relation to total expenditure made by households. The weight is kept constant by the INE along the 1993.02-2001.12 period, but since 2002 there has been a change in the methodology and the weights change every year. This fact is taken into account when the moments of the distribution of inflation are calculated. Another change in the methodology is the introduction of sales in the index. In order to avoid the problems caused by this change, we remove the seasonal component using the TRAMO-SEATS method. As control variables we use the rate of unemployment, the industrial production index, the general retail trade index, the shopping mall retail trade index, the oil prices and the industrial price index. We provide information about them in the corresponding section. As far as the construction of the main variables is concerned, we use the second and third cross-sectional moment of the distribution of price changes. The expressions of the standard deviation for each region (RPVjt) and the skewness for each region (Sjt) are as follows: [ ( ) ] 5,0 2 1 ∑ = −= n i jtijtijjt wRPV ππ ; [ ] ( ) 3 3 1 jt jtijt n i ij jt S w S ππ − = ∑ = where π refers to inflation rate, i to goods, j to regions and t to time periods. Therefore, π t: Spanish inflation in period t; π jt: inflation of region j in period t; π ijt: inflation of subgroup i in region j in period t; and wij is the weight of each subgroup i and region j used by INE. 3. Inflation, RPV and skewness: preliminary analysis on a regional basis In this section a preliminary region-by-region analysis is perfomed. In order to implement it, we run the following regression for each region Page 4 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 5 jtjt j jt j tj j jjt RPVS εββπβαπ ++++= −321,1 [1] The lagged inflation term is included in order to capture the persistence of the series. Before running the regressions we have checked the stationarity of the series.2 For the 17 regions inflation presents a negative deterministic trend for the 1993-1998 period, but there is no trend in the 1999-2005 one. This feature of inflation is included in the regressions. The regressions are estimated by ordinary least squares (OLS).3 As usual, the p-value of the t-statistic (in brackets in the tables) is corrected for heteroscedasticity by means of the White method. We show the results for each subperiod (Tables 2 and 3) and for the whole period (Table 4). [Table 2] [Table 3] [Table 4] As it can be seen from the tables, skewness is significant in 13 regions for the 1993-1998 subperiod, in 15 regions for the 1999-2005 subperiod, and in 13 regions for the whole period, and its coefficient remains unchanged for the different sample periods. However, the behaviour of RPV is not so homogeneous across periods, and tables show that it is significant in 7 regions for the first subperiod, it is not significant in any region for the second subperiod, and it is significant in 8 regions for the whole period (in 6 of them it was significant in the first period as well), and its coefficient varies considerably among sample periods. It is also interesting to point out the remarkable changes in the adjusted R2 depending on the period considered; the 2 In the Appendix we present the results for a common unit root –Breitung (2000) and Levin et al. (2002)-, and the general result is that it does not exist. Results of individual unit root tests are available from the authors upon request. The specific testing procedure adopted is the Augmented Dickey-Fuller (ADF) test with the Schwartz information criterion used to select the number of lags included in the ADF regressions. By default, the maximum number of lags allowed in the tests is 12. In the Appendix we also show the summary statistics for inflation, RPV and skewness. 3 As well known, if the lagged endogenous variable is not correlated with the error term, the validity of the OLS estimator holds. To prove that there is no correlation, we have estimated the model with OLS and verified that there is not autocorrelation in the residuals. Page 5 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 6 existence of a trend can be the key to this result. Finally, according to the coefficients on lagged inflation, it is clear that inflation shows persistence. In conclusion, these results seem to confirm the predictions of Ball and Mankiw’s model regarding the relevance of skewness, and show that RPV is more sensitive to changes in the inflation regime (the two sample periods in our analysis) than skewness. 4. Panel data analysis In this section, we perform panel data analysis in order to control for the possibility that regional inflation may be affected by common factors, which lead to strong correlation across regional inflation rates. In order to implement it, we attend to the following estimation: 17...1 321,1 = + + + + = −jRPVS jtjtjttjjjt ε β β π β α π [2] where αj is a fixed effect for each region. As it can be seen from equation (2), lagged inflation is correlated with the fixed effects. Therefore, within estimators will be biased and inconsistent. This problem cannot be avoided estimating the model in first differences, because although the fixed effect is wiped out, the first-differenced variables are correlated with the random component of the error term. The degrees of inconsistency and bias depend on T; only if T∞ the within estimator is unbiased and consistent.4 In other words, for a typical panel where N is large in relation to T (T is usually fixed), and where the enlargement of the sample always refers to N and not to T, instrumental variable estimation is required in order to get consistent and unbiased estimators. However, this is not our case because N (regions) is fixed, T is very large in relation to N, and the enlargement of the sample can be referred only to T. Despite the discussion about the number of periods required to get an unbiased and consistent within estimator would deserve a lot of attention, we have considered that the features of our sample allow us to use within estimators. 4 See Baltagi (1995, p. 126). Page 6 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 7 Now, we estimate (2) for the two subperiods5 and the total period –see Tables 5, 6 and 7, first columnand we perform a fixed effect test6 for the null hypothesis α j = α , for all j = 1…17. The test statistic is distributed under the null hypothesis as a F16,1169 and its value is 1.53 for the 1993-1998 period, as a F16,1390 and its value is 1.15 for the 1999-2005 period, and as a F16,2597 and its value is 1.04 for the total period. Therefore, the null hypothesis that α j are equal cannot be rejected in any case, so we estimate (3) –see Tables 5, 6 and 7, second column–: 17...1 321,1 = + + + + = −jRPVS jtjtjttjjt ε β β π β α π [3] Finally, the instrumental variable estimation suggested by Anderson and Hsiao (1981) is applied –see Tables 5, 6 and 7, third column. We estimate the model in first differences, in order to get rid of the hypothetical individual effects: )()()()( 1,1,31,22,1,11, −−−−−− −+−+−+−=− tjjttjjttjjttjtjtjjt RPVRPVSS εεββππβππ [4] As (πj,t-1πj,t-2) is correlated with the new error term, we run an instrumental variable estimation using the inflation variable in levels πj,t-2 as instrument; for the rest of the variables we do not define any instruments. [Table5] [Table 6] [Table 7] As it can be observed, there are not remarkable changes with respect to skewness for the three methods of estimation reported, and its coefficient seems to be stable across periods. But this does not hold for RPV and the constant term in the OLS estimation.7 These results lead us to introduce in the estimation for the total period both a dummy variable (D93-98) and a slope dummy (D93-98*RPVj,t) for the 1993-1998 period, in order to capture the change in the constant and in the coefficient of RPV respectively. Moreover, we have checked that a slope dummy for skewness is not 5 In order to reinforce the validity of the division in the sample period that we use in our analysis we have implemented a Chow test. The critical value of this test is 3.02 at 1% (the F statistic is 27.49) so we reject the null hypothesis of lack of a break in 1998:12. 6 See Baltagi (1995, p. 12). Page 7 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 8 significant. We have run the regression with fixed effects for the whole period, and again the Hausman test leads us to reject the fixed effects, so finally we present the results for the OLS estimation in Table 8: [Table 8] Summarising, our results show a homogeneous behaviour both across regions and periods regarding skewness, which can be revealing the vulnerability of the Spanish economy to supply shocks. As far as RPV is concerned, the predictions of Ball and Mankiw (1995) for no trend inflation are confirmed, given that it is not significant for the 1999-2005 period in any region. This variable appears to be heavily affected by the behaviour of the inflation rate. 5. Introduction of control variables As it was mentioned in the introduction, in this section we include several control variables. The idea embedded in the inclusion of these variables is twofold: i) to check the robustness of the aforementioned relation between mean inflation on the one hand and skewness and RPV on the other –Ball and Mankiw’s approach–; ii) to get some preliminary empirical evidence on the relevance of different macroeconomic relations for the Spanish economy. Although we have introduced many control variables, we would have liked to include even a higher number, but the monthly frequency imposed an important shortcoming (think for instance in variables related to fiscal policy). With the exception of the regional unemployment rate, these variables are provided at a national level, because they are nor available, homogeneously, at a regional level. The data sources for our control variables are the following8: i) Unemployment rates: Instituto Nacional de Empleo (INEM). ii) Industrial production index: INE (Base year 2000). iii) General retail trade index and shopping mall retail trade index: INE. iv) Interest rate: Bank of Spain. 7 Results for the constant term and trend are not included in the tables. They are available from the authors upon request. 8 A more detailed information about these variables and data sources is available from the authors upon request. Page 8 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 15 Quilis, E.M. (1999) Nota sobre la relación cíclica entre los índices de precios al consumo (IPC) e industriales (IPRI), Boletín Trimestral de Coyuntura, 73, 141-157. Raftai, A. (2004) Inflation and relative price asymmetry, European Central Bank, Working Paper Series, n. 301. Acknowledgments The authors would like to thank two anonymous referees, Carlos Dabús, Diego Romero-Avila, seminar participants at the Pablo de Olavide University and Centro de Estudios Andaluces, and participants at the EEFS 5th Annual Conference (Crete, Greece, 2006) and ERSA 46th Congress (Volos, Greece, 2006) for valuable comments and suggestions. The authors also acknowledge financial support from Junta de Andalucía: Centro de Estudios Andaluces (ECO 17-2004 and ECOD1.05/033) and CICE (Proyecto de Excelencia 01252). The usual disclaimer applies. Page 15 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 16 Table 1: Empirical evidence Authors Country and period Data Inflation regimes Main results *Ball and Mankiw (1995) US 1949-1989 Annual data Producer price Π around 3% Positive inflation/skewness association stronger than inflation/RPV Lourenco and Gruen (1995) Australia 1970-1992 Quarterly data Consumer and producer price L.I.P.: Π < 4-5% H.I.P.: Π > 4-5% The inflation/skewness relation stronger than inflation/RPV in L.I.P. The opposite is true for H.I.P. *Amano and Macklem (1997) Canada 1962-1994 Annual and quarterly data Producer price Low and stable inflation Positive inflation/skewness association and weak inflation/RPV association Hall and Yates (1998) UK 1975-1996 Monthly data Retail and producer price Changing inflation rate: Π > 12% in midseventies, negative rates in 1986-87 and early nineties For the whole period: inflation/skewness association is weaker than inflation/RPV Aucremanne et al. (2002) Belgium 1976-2000 Monthly data Consumer price H.I.P.: 1976-87, Π = 5,3% L.I.P.: 1987-00, Π = 2,8% Positive inflation/skewness and inflation/RPV associations independently of mean inflation *Döpke and Pierdzioch (2003) Germany 1969-2000 Annual data Consumer and producer price Changing inflation rate Similar positive inflation/skewness and inflation/RPV associations Assarson (2004) Sweden 1980-2002 Monthly and quarterly data Consumer price H.I.P.:1980-89 L.I.P.: 1990-02 RPV and skewness are more important in explaining inflation in the L.I.P. than in the H.I.P. Caraballo and Usabiaga (2004a) Spanish regions 1994-2001 Monthly data Consumer price Π < 5% Positive inflation/skewness association is stronger than the inflation/RPV one Caraballo and Usabiaga (2004b) Spain 1993-2001 Monthly data Consumer and Producer price Π < 5% Skewness is significant while RPV is not significant Raftai (2004) Hungary 1992-1997 Monthly data Consumer price Π: 15%-30% Positive inflation/skewness association Caraballo and Dabús (2005) Spain 1975-2002 Argentina 1960-1989 Monthly data Spain: Producer price Argentina: Wholesale price Spain: H.I.P.: 1975-85, Π: 14% L.I.P.: 1986-01, Π: 2,2% Argentina: L.I.P.: 1960-75, Π: 23%, H.I.P.: 1976-01, Π: 162% For both countries: H.I.P.: RPV is significant while skewness is not significant. L.I.P.: Skewness is significant and RPV only for Argentina. Π refers to the mean annual inflation rate, L.I.P. to low inflation period and H.I.P. to high inflation period. The asterisk implies that the work does not take into account the effects of inflation regimes on the results. Page 16 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 17 Table 2: Regional analysis (1993-1998) Region Constant π ππ π j,t-1 Sj,t RPVj,t Trend Adjusted R2 Andalucía 0.15 (0.00) 0.47 (0.00) 0.03 (0.00) 0.05 (0.02) -0.002 (0.00) 0.71 Aragón 0.29 (0.00) 0.40 (0.00) 0.02 (0.00) -0.01 (0.48) -0.003 (0.00) 0.75 Asturias 0.14 (0.35) 0.50 (0.00) 0.01 (0.00) 0.04 (0.57) -0.002 (0.00) 0.75 Baleares 0.19 (0.00) 0.26 (0.02) 0.004 (0.02) 0.08 (0.01) -0.002 (0.00) 0.83 Canarias 0.32 (0.00) 0.01 (0.83) 0.04 (0.00) 0.04 (0.09) -0.003 (0.00) 0.51 Cantabria 0.50 (0.00) -0.33 (0.00) 0.00 (0.95) 0.02 (0.35) -0.004 (0.00) 0.47 Cataluña 0.30 (0.00) 0.27 (0.01) 0.01 (0.00) 0.007 (0.86) -0.003 (0.00) 0.60 Castilla-León 0.30 (0.00) 0.19 (0.08) 0.005 (0.01) 0.02 (0.44) -0.004 (0.00) 0.64 Castilla-La Mancha -0.27 (0.07) 0.43 (0.00) 0.01 (0.00) 0.27 (0.00) -0.000 (0.59) 0.76 Extremadura 0.08 (0.24) 0.44 (0.00) 0.03 (0.00) 0.09 (0.00) -0.002 (0.00) 0.77 Galicia 0.28 (0.00) 0.36 (0.00) 0.01 (0.00) 0.00 (0.84) -0.003 (0.00) 0.78 Madrid 0.28 (0.00) 0.25 (0.02) 0.01 (0.00) 0.01 (0.70) -0.003 (0.00) 0.61 Murcia 0.23 (0.00) -0.37 (0.00) 0.01 (0.00) 0.18 (0.00) -0.004 (0.00) 0.63 Navarra 0.53 (0.01) 0.23 (0.04) 0.007 (0.08) -0.06 (0.56) -0.005 (0.00) 0.63 País Vasco 0.40 (0.00) -0.18 (0.14) 0.00 (0.21) 0.07 (0.05) -0.004 (0.00) 0.61 La Rioja 0.26 (0.17) -0.15 (0.22) 0.00 (0.87) 0.12 (0.17) -0.003 (0.00) 0.35 Valencia 0.05 (0.52) 0.35 (0.00) 0.03 (0.01) 0.12 (0.01) -0.001 (0.00) 0.60 Page 17 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 18 Table 3: Regional analysis (1999-2005) Region Constant π ππ π j,t-1 Sj,t RPVj,t Adjusted R2 Andalucía 0.14 (0.00) 0.39 (0.00) 0.03 (0.00) 0.003 (0.89) 0.26 Aragón 0.08 (0.28) 0.38 (0.00) 0.02 (0.00) 0.04 (0.30) 0.23 Asturias 0.11 (0.07) 0.42 (0.00) 0.01 (0.00) 0.01 (0.66) 0.24 Baleares 0.17 (0.00) 0.56 (0.00) 0.00 (0.00) -0.04 (0.06) 0.47 Canarias 0.28 (0.00) 0.26 (0.00) 0.02 (0.00) -0.10 (0.06) 0.16 Cantabria 0.35 (0.00) -0.52 (0.00) 0.01 (0.19) 0.01 (0.46) 0.24 Cataluña 0.20 (0.00) 0.21 (0.04) 0.01 (0.00) 0.01 (0.75) 0.09 Castilla-León 0.12 (0.09) 0.37 (0.00) 0.009 (0.02) 0.02 (0.54) 0.18 Castilla-La Mancha 0.09 (0.40) 0.37 (0.00) 0.02 (0.00) 0.03 (0.58) 0.22 Extremadura 0.05 (0.29) 0.42 (0.00) 0.03 (0.00) 0.03 (0.14) 0.27 Galicia 0.12 (0.02) 0.40 (0.00) 0.009 (0.02) 0.01 (0.50) 0.20 Madrid 0.14 (0.08) 0.21 (0.03) 0.01 (0.00) 0.03 (0.48) 0.18 Murcia 0.25 (0.00) -0.02 (0.78) 0.01 (0.00) 0.02 (0.44) 0.10 Navarra 0.17 (0.03) 0.20 (0.05) 0.01 (0.00) 0.02 (0.60) 0.09 País Vasco 0.27 (0.00) -0.01 (0.90) 0.006 (0.02) -0.001 (0.92) 0.02 La Rioja 0.31 (0.00) 0.01 (0.88) 0.009 (0.04) -0.001 (0.44) 0.02 Valencia 0.14 (0.03) 0.44 (0.00) 0.02 (0.16) 0.001 (0.97) 0.20 Page 18 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 19 Table 4: Regional analysis (1993-2005) Region Constant π ππ π j,t-1 Sj,t RPVj,t Trend (93-98) Adjusted R2 Andalucía 0.11 (0.00) 0.52 (0.00) 0.03 (0.00) 0.03 (0.04) -0.001 (0.02) 0.47 Aragón 0.10 (0.03) 0.46 (0.00) 0.01 (0.00) 0.04 (0.06) -0.001 (0.01) 0.43 Asturias 0.12 (0.01) 0.49 (0.00) 0.01 (0.00) 0.03 (0.09) -0.001 (0.00) 0.46 Baleares 0.07 (0.35) 0.71 (0.00) 0.005 (0.14) 0.02 (0.60) -0.000 (0.34) 0.70 Canarias 0.30 (0.00) 0.11 (0.02) 0.03 (0.00) 0.01 (0.35) -0.002 (0.00) 0.36 Cantabria 0.43 (0.00) -0.33 (0.00) 0.004 (0.53) 0.04 (0.06) -0.001 (0.00) 0.22 Cataluña 0.17 (0.00) 0.34 (0.00) 0.01 (0.00) 0.05 (0.06) -0.001 (0.00) 0.28 Castilla-León 0.10 (0.06) 0.45 (0.00) 0.005 (0.01) 0.05 (0.06) -0.001 (0.02) 0.35 Castilla-La Mancha -0.01 (0.77) 0.44 (0.00) 0.01 (0.00) 0.13 (0.00) -0.001 (0.01) 0.47 Extremadura 0.04 (0.39) 0.52 (0.00) 0.03 (0.00) 0.07 (0.00) -0.001 (0.01) 0.56 Galicia 0.14 (0.00) 0.48 (0.00) 0.01 (0.00) 0.03 (0.09) -0.001 (0.01) 0.46 Madrid 0.08 (0.15) 0.33 (0.00) 0.01 (0.00) 0.09 (0.00) -0.001 (0.01) 0.32 Murcia 0.22 (0.00) -0.01 (0.88) 0.01 (0.00) 0.08 (0.00) -0.001 (0.00) 0.26 Navarra 0.20 (0.00) 0.29 (0.00) 0.01 (0.00) 0.07 (0.01) -0.002 (0.00) 0.34 País Vasco 0.27 (0.00) 0.12 (0.30) 0.004 (0.15) 0.04 (0.05) -0.001 (0.00) 0.32 La Rioja 0.32 (0.00) 0.05 (0.52) 0.002 (0.43) 0.02 (0.17) -0.001 (0.00) 0.12 Valencia 0.08 (0.07) 0.48 (0.00) 0.03 (0.01) 0.05 (0.05) -0.000 (0.07) 0.37 Page 19 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 20 Table 5: Panel data analysis (1993-1998), with negative trend Variable Fixed Effect OLS Anderson-Hsiao π j,t-1 0.12 (0.00) 0.25 (0.00) 0.12 (0.00) Sj,t 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) RPVj,t 0.05 (0.00) 0.04 (0.00) 0.006 (0.03) Adjusted R2 0.58 0.57 - Table 6: Panel data analysis (1999-2005) Variable Fixed Effect OLS Anderson-Hsiao π j,t-1 0.28 (0.00) 0.29 (0.00) 0.57 (0.00) Sj,t 0.01 (0.00) 0.01 (0.00) 0.006 (0.00) RPVj,t 0.01 (0.25) 0.01 (0.05) 0.00 (0.7) Adjusted R2 0.17 0.16 - Table 7: Panel data analysis (1993-2005), with negative trend (1993-1998) Variable Fixed Effect OLS Anderson-Hsiao π j,t-1 0.35 (0.00) 0.36 (0.00) 0.39 (0.00) Sj,t 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) RPVj,t 0.04 (0.00) 0.04 (0.00) 0.002 (0.23) Adjusted R2 0.32 0.34 - Table 8: Panel data analysis with dummies (1993-2005). OLS Constant πj,t-1 Sj,t RPVj,t D93-98*RPVj,t D93-98 Trend (93-98) Adjusted R2 0.38 (0.00) 0.28 (0.00) 0.01 (0.00) 0.01 (0.02) 0.02 (0.01) -0.14 (0.00) -0.003 (0.00) 0.38 Page 20 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 21 Table 9: Introduction of control variables: unemployment, industrial production, retail trade, interest rates, oil prices and industrial price index Constant 0.38 (0.00) 0.38 (0.00) 0.39 (0.00) 0.38 (0.00) 0.37 (0.00) 0.44 (0.00) 0.35 (0.00) 0.35 (0.00) πj,t-1 0.28 (0.00) 0.28 (0.00) 0.27 (0.00) 0.27 (0.00) 0.27 (0.00) 0.27 (0.00) 0.31 (0.00) 0.26 (0.00) Sj,t 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) 0.01 (0.00) RPVj,t 0.01 (0.02) 0.01 (0.02) 0.01 (0.02) 0.01 (0.02) 0.01 (0.00) 0.01 (0.04) 0.01 (0.03) 0.01 (0.09) D93-98*RPVj,t 0.02 (0.01) 0.02 (0.01) 0.02 (0.02) 0.02 (0.00) 0.02 (0.06) 0.01 (0.17) 0.02 (0.01) 0.02 (0.01) D93-98 -0.14 (0.00) -0.14 (0.00) -0.14 (0.00) -0.14 (0.00) -0.12 (0.00) -0.14 (0.00) -0.12 (0.00) -0.12 (0.00) Trend (93-98) -0.003 (0.00) -0.003 (0.00) -0.003 (0.00) -0.003 (0.00) -0.003 (0.00) -0.004 (0.00) -0.002 (0.00) -0.002 (0.00) Spanish cyclical unemployment -0.006 (0.50) Regional cyclical unemployment -0.002 (0.69) Cyclical industrial production index 0.002 (0.01) Cyclical general retail trade index 0.01 (0.00) Cyclical shopping mall retail trade index 0.008 (0.00) Lagged (t-15) change in interest rates -0.11 (0.02) Change in oil prices 0.30 (0.00) Lagged (t-3) change in industrial price index 0.06 (0.00) Adjusted R2 0.38 0.38 0.39 0.39 0.39 0.34 0.42 0.39 Page 21 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For Peer Review 22 Appendix Table A1: Panel data unit root analysis (1993-1998, with trend) and summary statistics Variable Levin, Lin and Chu (2002) Breitung (2000) Statistic Prob. Statistic Prob. Sj,t -30.62 0.00 -17.84 0.00 RPVj,t -14.29 0.00 1.58 0.00 π j,t -12.18 0.00 -3.70 0.00 Mean Max. Min. Sj,t 0.58 12.98 -9.84 RPVj,t 1.46 3.56 0.52 π j,t 0.28 0.81 -0.22 Table A2: Panel data unit root analysis (1999-2005) and summary statistics Variable Levin, Lin and Chu (2002) Breitung (2000) Statistic Prob. Statistic Prob. Sj,t -25.50 0.00 -13.98 0.00 RPVj,t -4.29 0.00 -2.82 0.00 π j,t -27.57 0.00 -18.19 0.00 Mean Max. Min. Sj,t 0.43 10.48 -12.48 RPVj,t 1.62 2.88 0.53 π j,t 0.26 0.74 -0.27 Table A3: Panel data unit root analysis (1993-2005, with trend) and summary statistics Variable Levin, Lin and Chu (2002) Breitung (2000) Statistic Prob. Statistic Prob. Sj,t -48.38 0.00 -26.92 0.00 RPVj,t -4.20 0.00 0.42 0.66 π j,t -24.53 0.00 -5.86 0.00 Mean Max. Min. Sj,t 0.50 12.90 -12.21 RPVj,t 1.55 3.56 0.52 π j,t 0.27 0.81 -0.27 Page 22 of 22 Editorial Office, Dept of Economics, Warwick University, Coventry CV4 7AL, UK Submitted Manuscript 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60