1. INTRODUCCIÓN
1.1. El cultivo bajo invernadero en la zona mediterránea
1.1.2. Sustratos utilizados en la horticultura intensiva
1.1.2.2. Factores que limitan la absorción radicular de
We apply the model to the UK data. The objective is to contrast and compare the use of macroeconomic measurements in credit card model with two datasets to make sure that some properties of the models might hold in general while others depend on the particular economy under investigation. We will not discuss the estimates in depth and instead only present the critical estimates. The full results are presented in the Appendix B.
We present the estimates of the multinomial and cumulative logistic regression coeffi- cient estimates of the UK dataset in Tables 5.14, B.3, 5.15 and B.4.
We first compare the results in Table 5.14 with those in Table 5.5. For the CPI model, the implications for the HK and UK data are ”‘A > T > C”’ and ”‘A > C > T”’ for inactive accounts respectively. These indicate that when the economy experiences inflation, more Inactive accounts activate their credit cards. On the other hands, when there is inflation, the Others accounts in HK and UK react differently. In HK, more Others borrowers reduce the use of their credit cards since the implication is ”‘T > A > C”’. The UK borrowers react more extremely to inflation as many of them close their accounts (since the implication is ”‘C > A > T”’) when there is inflation. The increase of interest rate is also an indicator for inflation and therefore the results in the Interest Rate model
Scorei Macroeconomics Format Log(A/C) Log(T/C) Log(A/T) -2Log(likelihood) Implications Measurement (k,w) β β β Inactive CPI (9,1) 1.9152* -9.3047* 11.2199* 222005 A>C>T Risk CPI (N,N) N / / - Others CPI (12,.8) -2.333* -3.6692* 1.3362* 255310 C>A>T Inactive GDP (6,1) -3.436* -0.3594* -3.0765* 219995 T>C>A Risk GDP (12,.8) 0.585** / / 1592 A>C Others GDP (9,1) -0.2084* 0.3474* -0.5558* 255449 T>C>A Inactive Int (1,1) 6.5946* 6.1976* 0.397** 228573 A>T>C Risk Int (9,1) -9.5591** / / 1589 C>A Others Int (12,1) -10.5283* -7.0736* -3.4547* 255240 C>T>A Inactive Sto (2,.8) -18.5047* -26.1677* 7.663* 223946 C>A>T Risk Sto (2,0.2) -5.8758** / / 1586 C>A Others Sto (3,.8) 3.8009* 3.5201* 0.2809 255660 A>T>C Inactive Une (1,1) -0.7448* 3.3103* -4.0551* 219815 T>C>A Risk Une (9,1) -9.5591** / / 1589 C>A Others Une (12,1) -10.5283* -7.0736* -3.4547* 255240 C>T>A ”/” represents there is no observation in the data.
”‘N”’ represents the stepwise multinomial logistic regression cannot find any significant explanatory macroeconomic variable.
”‘*”’ indicates the parameter is significant at 99% level. ”‘**”’ indicates the parameter is significant at 95% level.
The first column is the index of the initial score stateiwhere ”‘Others”’ refers to accounts with ordinary behavioural score (Score1 to Score4).
The best fit macroeconomic variables (discussed in Section 5.2.4) are presented in column three. The estimated parameters are presented in column four to six.
-2log(likelihood) ratios which are used to measure the model fit statistics are presented in column seven.
Log(A/C) represents log
p(A|i,M) p(C|i,M)
in (5.2) Log(T/C) represents log
p(T|i,M) p(C|i,M)
in (5.3) Log(A/T) represents logpp((AT||i,i,MM))in (5.4)
Table 5.14: Summary of the multinomial logistic model estimates for the UK dataset
are very similar to those of the CPI model. When the interest rate goes up, more Inactive borrowers in HK or UK require additional credit since more of them are moving to an active status. The Others borrowers in UK tend to close their account whenever the interest rate increases but those in HK tend to remain inactive.
The way HK and UK Inactive borrowers react to GDP is very similar. When GDP goes up (i.e. the economy is doing well), more HK Inactive borrowers close their accounts (since the implication is ”‘C > T > A”’) and more UK Inactive borrowers remain inactive (since the implication is ”‘T > C > A”’). In other words, the Inactive borrowers do not need credit when the economy is doing well. For the Others borrowers, the implications of the HK and UK data are ”‘T > A > C”’ and ”‘T > C > A”’ indicating more of them move to an Inactive status. So in summary, these results show borrowers do not want credit during good times.
The borrowers’ reaction to Stock market is different in the HK and UK markets. When there is bull market, more Inactive borrowers in HK activate their credit card immediately (since the lag=1 and the implication is ”‘A > T”’). Conversely, in UK, more Inactive borrowers close their accounts when the stock market is doing well. For the Others borrowers in HK, they tend to reduce borrowing with their credit card when the Hang Seng index increases, whereas those in UK keep their current status unchanged (i.e. remain Active).
The reaction to the labour market is quite similar in these two credit card datasets. The HK borrowers, either the Inactive or Others, close their credit cards when the un- employment rate increases. In the UK market, Inactive borrowers remain as inactive but Other borrowers tend to close their credit card account when the unemployment rate goes up. We performed multinomial analysis for Model A and Model B and the results are presented in the Appendix B for reference.
Initiali CPI GDP Interest rate Stock Unemployment (k,w) β -2Log(L) (k,w) β -2Log(L) (k,w) β -2Log(L) (k,w) β -2Log(L) (k,w) β -2Log(L) Inactive (6,1) -1.0618** 133333 (3,1) -0.2249** 13332 (3,1) -1.3373** 13340 (2,0.2) -1.6606** 13337 (3,1) -1.3373** 13340 Risk (9,1) 1.5535** 7382 (9,1) 0.7646* 7354 (12,1) 6.8403** 7376 (3,0.8) 3.286** 7380 (12,1) 6.8403** 7376 Score1 (1,1) 0.0881* 385808 (12,1) -0.3608* 385442 (6,0.8) 0.8759* 385830 (1,1) -0.8003* 385797 (6,0.8) 0.8759* 385830 Score2 (1,1) -0.1732* 559078 (9,1) -0.2894* 558900 (12,1) -2.5087* 559150 (1,1) -1.2616* 559112 (12,1) -2.5087* 559150 Score3 (1,1) -0.2399* 604810 (9,1) -0.185* 605137 (1,1) 0.6274* 605191 (2,0.5) -1.4517* 605161 (1,1) 0.6274* 605191 Score4 (1,1) -0.3256* 364450 (12,1) 0.7598* 363388 (9,1) -5.9774* 364486 (3,0.8) 3.9488* 364727 (9,1) -5.9774* 364486
”‘*”’ indicates the parameter is significant at 99% level. ”‘**”’ indicates the parameter is significant at 95% level.
Table 5.15: Summary of cumulative logistic model estimates for the UK dataset
We look at the results in Table 5.7 and 5.15 to compare the behavioural score migration of credit card borrowers in HK and UK. Since most of the coefficient estimates with respect to the CPI model in both markets are negative, these datasets show borrowers’ behavioural score increases when there is inflation. All coefficient estimates with respect to the GDP model are negative. These indicate that the behavioural score of borrowers increases when the economy is doing well (which is shown by the increase in GDP).
The borrowers in HK and UK react differently with respect to increasing Interest Rate. Most of the coefficient estimates of the Interest Rate model of the UK dataset are positive. These indicate that borrowers’ behaivoural score reduces when there is inflation. Conversely, the behavioural score of the HK borrowers increases when the Interest Rate goes up since all coefficient estimates of the Interest Rate model of the HK dataset are negative.
Most of the coefficient estimates with respect to the Stock market model in both markets are negative. This indicates the behavioural score of borrowers improves when there is bull market. One surprising result of the Unemployment model in UK is that there are negative coefficient estimates. These indicate borrowers’ move to a state with high behavioural score when the unemployment rate increases. Note, however, that the labour market in UK was rather stable over the sampling period and thus this result might not reflect the actual impact of unemployment rate on the UK market.