Supporting information:
Historical data (Excel spreadsheet)Survey notes for tables C12 and C13
Chronology of changes to tables and current status
January 2012: Improvements to seasonally adjusted credit card billings series
The Rugby World Cup (RWC) in September and October 2011 attracted over 100,000 overseas visitors to New Zealand. This resulted in a significant increase in credit card billings in New Zealand by visitors using overseas issued credit cards. In this note we explain how we have accounted for the impact of the Rugby World Cup in the model we use to seasonally adjust credit card series and the impact on the previously published estimates.
Seasonal adjustment is the process of estimating and removing the usual seasonal effects from a time series in order to reveal the underlying non-seasonal features. The seasonal factors for a given month, say September, are influenced by the September months in other years. The RWC was a large one-off event that did not fit the usual seasonal pattern and to prevent it having an undue influence on the seasonal factors we have explicitly treated it as an outlier in the seasonal adjustment model. This results in a higher quality seasonally adjusted series of credit card billings in New Zealand on overseas issued cards.
The following graphs compare the previously published series of credit card billings on overseas issued cards (where there was no explicit treatment for the RWC in the seasonal adjustment model) and the new series (which explicitly accounts for the RWC). The main impact is a higher rate of growth in September and October 2011.
|
|
|
The model used to seasonally adjust the overseas credit card billings on New Zealand issued cards has also been updated to reflect changes caused by the later timing of the school holidays in 2011. The new series remains largely unchanged from the previously published series.
For more information on the model see the ‘Seasonal adjustment practices’.
September 2011: Improvement in credit card limit statistics
In August 2008 the credit card limit series was re-specified to include only limits on cards that are active (unblocked) and able to be used for spending; excluded from that month were cards that were not active (blocked) and unable to be used for spending. This change in definition resulted in a series break.
This month the credit limit series has been revised. A long-run historical unblocked credit limit series has been estimated to remove the break. A minor revision to the series since 2008 has also been included due to improved information from a respondent.
September 2010: New series added to table C12
This month the Reserve Bank introduced two new series to its credit card statistics.
The first is a 90 day overdue series. This is the value of advances outstanding (or credit card debt) that is older than 90 days. This time series enables analysis of the ability or willingness of credit card users to repay debt.
The second is a credit card limit utilisation series. Credit limit utilisation is calculated as the ratio of total advances outstanding (or credit card debt) to total allowable credit limits.
Changes have also been made to the seasonal adjustment of credit card series. See the ‘ Seasonal adjustment practices’.
Please note that in addition to the two new series, the other series in the Excel version of Table C12 have been rearranged.
June 2009: Revised personal credit card interest rate for table C12
This month the RBNZ published a revision to the weighted average interest rate (WAIR) series for interest-bearing advances from February 2009. In the past, credit card issuers charged the same interest rate to borrowers, whether they used a credit card for purchasing goods and services, or for obtaining cash. However, most issuers have recently introduced a higher interest rate for borrowing cash using a credit card. This development has resulted in a slight under-estimation of the interest-bearing advances WAIR. This month revisions have been made for three respondents. Minor revisions for other respondents can be expected in the following months.
August 2008: Series break in credit limits for table C12
This month a series break has occurred in the credit limits series in table C12. This is the result of the credit limits data being re-specified to include only limits on cards that are active (‘unblocked’) and able to be used for spending at the reference date. Excluded from the calculation of credit limits, from August 2008 onwards, are cards that are ‘blocked’ or not able to be used for spending at the reference date.
The level shift in the series does not represent any change in approach by credit card issuers to providing limits. For this reason annual growth rates are misleading and have been excluded from the published tables. These changes also impacted on the credit limit utilisation series being different prior to August 2008.
For more information see the background notes for the published credit card tables.
August 2008: Table C12 and C13 - Seasonal adjustment
Most data series collected in the credit card survey are affected by seasonal factors, for example national holidays and festivities, the timing of school holidays, and the number of trading days in the month. The existence of such seasonality complicates month-to-month comparisons of growth rates. To compensate for these seasonal differences, the Reserve Bank of New Zealand (RBNZ) used the US Census Bureau’s (1967) X11 software.
Beginning with the publication of the August 2008 monthly survey, the RBNZ is introducing the use of the US Census Bureau’s X-12-ARIMA (Release 0.3), a widely applied seasonal adjustment method. The X-12-ARIMA package is an addition to the X11 family and contains ARIMA modeling capabilities, a wider range of statistical diagnostics for quality control, and automatic detection and correction of outliers for all seasonally adjusted series. For detailed information on the technical capabilities of X-12-ARIMA, please view the US Census Bureau’s website. The seasonal adjustment specifications used are included in the special notes appendix to the historical data spreadsheets.
As a result of the migration to X-12-ARIMA, the revised seasonally adjusted data have exhibited improved end point stability and lesser volatility around the Easter holiday. The graph below compares results from X11 and X-12-ARIMA as applied on amounts financed in $NZ on domestic and overseas issued cards.
The presentation and layout of credit card survey data have also been reviewed to make the seasonally adjusted data accessible to users (see hC12.xls and hC13.xls). In a number of series the RBNZ’s published tables display year-on-year percentage changes based on the seasonally adjusted data, rather than on the actual series. Where components of a series display significantly different seasonal profiles, the RBNZ removes the seasonality before summing them to derive an adjusted total. Where seasonality is not significant, the seasonally adjusted series is set equal to the actual values over the indicated time span.
The seasonal adjustment procedure will be applied with each new set of monthly observations. Going forward the seasonal adjustment specifications will be reviewed on an annual basis to ensure they remain relevant.
A pre-defined adjustment model chosen on statistical merit is used for each series;
An adjustment is run on the entire series as each new observation becomes available (concurrent seasonal adjustment method);
Where a series is an aggregation of multiple series, if justified, its seasonally adjusted values are taken as the sum of each separately adjusted sub-series (indirect adjustment);
Each adjustment model is reviewed annually.
Adjustment models used
(1) Amounts financed on cards used in New Zealand (Domestic cards)
Transformation: Log form
(a) Dec-1993 to Dec-2000:
ARIMA (2,1,0)(0,1,1)12
Regression variables: Constant term, trading day and Easter
X11 multiplicative model, seasonal 5-term and Henderson 13 term MA
(b) Jan-2001 onwards:
ARIMA (0,1,2)(0,1,1)12
Regression variables: Constant term, trading day and Easter; additive outlier Dec 2001, transitory change Feb 2008
X11 multiplicative model, seasonal 5 & 7 terms and Henderson 15 term MA
(2) Amounts financed on cards used in New Zealand (Overseas cards)
Transformation: Log form
ARIMA (2,1,2)(0,1,1)12
Regression variables: Constant term, trading day, additive outliers (Oct-1998, Sep-2003, Sep-Nov 2011), transitory changes (Sep-2003, Jun-2005)
X11 multiplicative model, seasonal 5-term and Henderson 13 term MA
(3) Amounts financed on cards issued in New Zealand (Overseas billings)
Transformation: Log form
ARIMA (0,1,1)(0,1,1)12
Regression variables: Constant term, trading day, additive outlier (Jul-1997, Sep-Oct 2011)
X11 multiplicative model, seasonal 7-term and Henderson 15 term MA
(4) Amounts financed on cards used in New Zealand (Total billings in New Zealand)
as per (1) and (2) above
(5) Amounts financed on cards issued in New Zealand (Total billings on NZ cards)
Transformation: Log form
(a) Jan-1981 to Nov-1993:
ARIMA (2,1,0)(0,1,1)12
Regression variables: Constant term, trading day, Easter, level shift (Mar-1981)
X11 multiplicative model, seasonal 7-term and Henderson 13-term MA
(b) Dec-1993 onwards, as per (1) and (3) above
(6) Deposit balances
Transformation: Log form
ARIMA (0,1,1)(0,1,1)12
Regression variables: constant term, trading day, additive outliers (Apr-2003, Feb-2008)
X11 multiplicative model, seasonal 5 & 7-term and Henderson 9-term MA
(7) Total advances outstanding
(a) Jan-1981 to Jan-1985:
No seasonality
(b) Feb-1985 to Jun-2001:
Transformation: Log form
ARIMA (2,1,2)(0,1,1)12
Regression variables: Constant term, additive outliers (Sep-1999, Jan-2001), transitory change Sep-1988, level shift Dec-2000
X11 multiplicative model, seasonal 7-term MA and Henderson 15-term MA
(c) Jul-2001 onwards:
Transformation: Log form
ARIMA (0,1,1)(0,1,1)12
Regression variables: constant term, trading day
X11 multiplicative model, seasonal 7-term MA and Henderson 9-term MA
(8) Personal interest-bearing advances outstanding (data based on daily average balances for the month)
Transformation: Log form
ARIMA (2,1,2)(0,1,1)12
X11 multiplicative model, seasonal 5 & 7-term and Henderson 13 term MA
(9) Personal non-interest-bearing advances outstanding (data based on daily average balances for the month)
(a) Jul-2000 to Dec-2005:
No seasonality
(b) Jan-2006 onwards:
Transformation: Log form
ARIMA (0,1,1)(1,0,0)12
Regression variables: constant term, trading day, additive outliers (Feb-2008, Jan-2010)
X11 multiplicative model, seasonal 5 & 7-term MA and Henderson 13-term MA
(10) Total advances outstanding (data based on daily average balances for the month)
Summation of values from (8) and (9) above
October 2001: Series break for tables C12 and C13
A minor series break (less than 1.5% of totals in all series) has occurred with the introduction of another respondent to the survey.
March 2001: New credit card survey data for tables C12 and C13
Revised and additional data from the new credit card survey have been disseminated since March 2001, backdated to July 2000. New data series are available for C12 ‘Balances outstanding, limits and interest rates’. For C13 ‘Amounts outstanding’ (billings), the presentation of existing data series has been improved to provide more useful information. In both cases, minor revisions to past data have been made.
'Table C12 retains the long-run ‘total advances outstanding’ and total credit limit series, for which historical data is available from the website from 1981. The long-run series previously labeled ‘weighted average interest rate’ reported the interest rates of the ‘classic’ most commonly available card, weighted by total outstandings of each issuer. This series has been recalculated as a simple average. Until August 2000, there is very little change to results obtained using the weighted average, but from this point two new classic card issuers’ rates are introduced (AMEX and TSB Bank), which accounts for the break in the new simple average series at that point. All of this data has a reference date of the end of the month. From and including October 2001, The Warehouse data have been included in the survey, affecting all but one of the C12 data categories to an extent.
New data in C12 includes deposit balances held in credit card accounts at month end. In addition, for personal cards only, on a daily average basis the monthly total advances outstanding is split between interest-bearing and non-interest-bearing balances. This was formerly estimated at month end, on all balances (business and personal). As part of this change, a weighted average interest rate on interest-bearing balances has been calculated using approximation techniques explained in the series description. The effective rate of interest on all personal balances is also provided. By way of comparison, the December 2000 ‘weighted average interest rate’ for classic cards only was 19.2% - weighted by all cards on issue the interest rate is 18.1%. The extent of this difference is no longer as marked when measured against the new simple average series because the two new classic card issuers’ card rates introduced to the new historical simple average series are lower than the average of the five card rates used under the former methodology.
'Table C13, formerly titled ‘Monthly billings’, provides the three existing series in two different formats. Minor revisions have been made to the data, which is available in the historical format from 1981 for billings on domestic cards, and from December 1993 for billings overseas on domestic cards and in New Zealand on cards issued overseas. The former total billings series is now ‘amounts financed on cards issued in New Zealand’, arrived at by combining domestic card billings in New Zealand and overseas: this is the series that began in 1981. A new total is computed for amounts financed on cards used in New Zealand, by adding billings in New Zealand on domestic cards to spending on cards in New Zealand by visitors to New Zealand. From and including October 2001, there is a change in coverage, causing a series break, affecting all but one of the C13 data categories to an extent.
Note that as before, rates of growth of billings are calculated on a seasonally-adjusted basis, taking into account trading days.
August 2000: Survey coverage increased for table C12
For personal cards only, new data in C12 includes deposit balances held on cards, and on a daily average basis the monthly total advances outstanding split between interest-bearing and non-interest-bearing balances. As part of this change, a weighted average interest rate on interest-bearing balances has been calculated using approximation techniques explained in the series description. The effective rate of interest on all personal balances is also provided. In addition, the long-run series previously labelled ‘weighted average interest rate’ reported the interest rates of the ‘classic’ most commonly available card, weighted by total outstandings of each issuer. This series has been recalculated as a simple average. Until August 2000, there is very little change to results obtained using the weighted average, but from this point two new classic card issuers’ rates are introduced, which accounts for the break in the new simple average series at that point. All of this data has a reference date of the end of the month.
December 1993: Revised amounts financed for table C13
Table C13 has been revised to provide three series for amounts financed. Minor revisions have been made to the data available from 1981 for billings on cards issued in New Zealand. From December 1993 this series was split to provide billings overseas on domestic cards, and amounts financed on domestic cards in NZ. The former total billings series is now ‘amounts financed on cards issued in New Zealand’, arrived at by combining domestic billings on cards issued in New Zealand and overseas billings on these cards. From December 2003, a new billings series was introduced for spending in NZ on cards issued overseas. A new total is computed for amounts financed on cards used in New Zealand, by adding billings in New Zealand on domestic cards to spending on cards in New Zealand by visitors to New Zealand.
June 1992: Survey coverage increased for table C12
Table C12 introduced an end of month simple average standard card interest rate. This is based on the interest rate of the ‘classic’ most commonly available card reported for each issuer.
April 1991: Survey coverage increased for table C12
A total limit on credit cards was added, followed by an end of month simple average standard card interest rate. Billings on domestic cards and for billings overseas on domestic cards and in New Zealand on cards issued overseas were added in December 1993. Finally in July -and August 2000 the tables were expanded to reflect the current format of the survey.
December 1993: Survey coverage increase for tables C12 and C13
Table C12 retains the long-run ‘total advances outstanding’ which historical data is available from the website from 1981. Table C13 the total billings series, began in 1981.The data for the credit card survey is available from the website, consisting mainly of total advances outstanding as well as total billings on cards issued in New Zealand.
