Don't buy a stock until you know its potential and set a target price.
A simple, 5-step process, using Oracle as an example.
By Harry Domash
Most money managers wouldn't consider buying a stock until they've set a target price, and neither should you.
Your target price is the price you think a stock will hit at a specified future date.
I'm going to describe a method that you can use to calculate target prices that is deceptively simple. It uses historical data rather than guidance from the companies themselves or Wall Street analysts. It involves forecasting a company's sales per share, and then using its historical price-to-sales ratios to set target stock prices.
To demonstrate the process, I'll compute a target price and tell you whether it's worth investing in software applications supplier Oracle
Why use sales and price-to-sales instead of earnings and price-to-earnings ratios? First, sales growth is easier to predict than earnings growth. Plus, if you look at historical data, you'll find that P/E ratios are a lot more volatile -- and thus, harder to forecast -- than P/S ratios.
Since the process is based on forecasts, and forecasts are always wrong, I won't try to set a precise target price. Instead, I'll estimate a low and high target-price range.
A five-step process
The stipulation is that the target-price date is always the day after a company reports its fiscal-year results. I call that fiscal year the target year.
Developing my target price consists of five steps:
• Estimate sales in the target year.
• Estimate the number of shares outstanding in the target year.
• Use the results from steps 1 and 2 to compute estimated target-year sales per share.
• Estimate expected range of price/sale ratios.
• Use No. 3 and No. 4 to compute the estimated target price range.
Now, we'll set a target for Oracle.
The company's fiscal year ends in May, so I'll use its 2007 fiscal year. Oracle will probably report its May 2007 fiscal year results in June or July 2007.
I've found that the calculations go faster if I first print MSN's key ratios 10-year summary and financial statements 10-year summary reports for stocks I analyze.
Step 1: Start with sales
Start by estimating a company's target fiscal-year sales. The 10-year financial statements summary shows each company's fiscal-year sales going back 10 years. Most analysts forecast sales growth in terms of year-over-year percentage increase. However, I've found that it's more useful to look at recent historical sales growth in terms of actual dollars instead of percentages.
Oracle's recent sales growth has been volatile, ranging from a $1.3 billion year-over-year gain in fiscal 2000 to a $1.2 billion drop in fiscal 2002. In its most recent fiscal year, ending in May 2004, sales climbed $681 million. I calculated Oracle's five-year average growth at $266 million, which isn't much compared to its $10.2 billion fiscal-year 2004 sales total.
Starting with Oracle's 2004 sales of $10.156 billion, I added $266 million to get $10.422 billion for 2005. Adding another $266 million results in $10.688 billion for its May 2006 fiscal year.
Finally, adding $266 million to that figure yields estimated sales of $10.954 billion for its May 2007 target year.
• Oracle target year sales: $10.954 billion
My sales estimate assumes that recent annual historical sales growth will continue. Obviously, that's not always the case. So modify your target-year sales if you have more reliable numbers.
Step 2: Shares outstanding
Next, I estimate a company's total shares outstanding at the end of its target year. Again, I use history as my guide. Many companies consistently increase their number of shares outstanding as they issue stock to raise cash, make acquisitions or allocate shares for employee stock options.
Oracle has reduced its number of shares outstanding in recent years. On average, its total dropped by 100 million shares annually over its past five fiscal years. Using that figure, I estimated that Oracle's average 5.2 billion shares outstanding in 2004 would drop to 4.9 billion by its fiscal 2007 target year.
• Oracle target year shares outstanding: 4.9 billion
Step 3: Sales per Share
Just as earnings per share is annual earnings divided by the number of shares outstanding, sales per share is annual sales divided by the number of shares out.
I estimated Oracle's target-year sales of $10.954 billion and shares outstanding at 4.9 billion. So my estimated target year sales per share ($10.954 divided by 4.9, rounded down) is $2.20.
• Oracle sales per share: $2.20
Step 4: Price/sales ratios
Investors frequently compare valuation ratios to evaluate the relative merits of companies in the same industry. For instance, Company A is the best buy if its P/E is only 20, while Company B's P/E is 35.
Competing firms often consistently trade at different valuations depending on their popularity with investors. For instance, pharmaceutical maker Pfizer (PFE, news, msgs) almost always trades at higher valuations than competitor Merck (MRK, news, msgs). This is true no matter what valuation ratio you choose. In terms of price/sales ratios, Pfizer's 6.6 average P/S over the past five years is almost double Merck's 3.5 figure. Similarly, chip maker Intel (INTC, news, msgs) has traded at an average 5.8 P/S over the past five years compared to 1.3 for competitor Advanced Micro Devices (AMD, news, msgs).
Thus, instead of comparing valuation to the overall market or to its sector, I've found that a stock's own history is the best indicator of its likely future trading ranges.
MSN's Key Ratios report shows the average annual price/sales ratios going back 10 years. I think the most recent five years' data are the most relevant, and that's what I use to determine the range of anticipated P/S ratios at my target date.
Over nine of the past 10 years, Oracle has traded at P/S ratios ranging from 4 to 7.9. However, in its May 2000 fiscal year, MSN lists its average P/S at 20. It's best to ignore an obviously out-of-range figure.
Disregarding the 2000 figure, Oracle's last five P/S ratios ranged from 4 to 7.9.
• Oracle target P/S: 4 to 7.9
Step 5: Doing the numbers
If you remember your algebra, you'll know that share price can be calculated by multiplying the sales per share by the price-to-sales ratio. For instance, a stock would be trading at $20 if its sales per share were $10, and the P/S ratio was 2 (it works: price/sales = 20/10 = 2).
Target Price = sales x P/S
In step 3, I estimated that Oracle would have sales of $2.20 per share. In step 4, I estimated its price/sales range at 4 to 7.9. Multiplying by sales per share by P/S:
• Oracle target price range: $8.80 to $17.40.
Oracle recently changed hands at $11.70; already within my $8.80 to $17.40 summer 2007 estimated trading range. I'd abandon Oracle for better prospects.
This simple target-price calculation is intended to help you evaluate stocks that you're researching. But it doesn't take changing economic or competitive conditions into account. In short, it's no substitute for doing your own due diligence.
Domash publishes the Winning Investing stock and mutual fund advisory newsletter and writes the online investing column for the San Francisco Chronicle. Harry has two investing books out, the most recent being "Fire Your Stock Analyst," published by Financial Times Prentice Hall.
Source: http://articles.moneycentral.msn.com/Investing/SimpleStrategies/FindStocksWithMoreProfitPotential.aspx
Tuesday, October 16, 2007
APPLICATION OF MARKOWITZ PORTFOLIO ANALYSIS IN INDIAN STOCK MARKET
description of application of Markowitz portfolio analysis to a group of shares recomended by a broker (India)
Sources of Data:
Valueline is a monthly bulletin published by Sharekhan (2005) a broking firm in India. The bulletin contains the target price information and the market price on the date of publication for various stocks researched and recommended by the firm. The data from the bulletin of July 2005, which was made available on the website of the firm for public access, is selected for getting the data of expected returns. Target price data was available for 43 companies. Covariance is to be calculated using 25 months closing price data. The monthly closing price data was taken from Prowess, an electronic data base of balance sheet and share price data of Indian companies published by Centre for Monitoring Indian Economy (CMIE, Mumbai). Out of the total 43 companies, for two companies, data was not available for the full 25 months. These two companies were dropped from the set of securities considered for forming the portfolio. Hence, the final list of stocks considered for portfolio analysis contains 41 companies.
Calculation of Input Variables:
The expected returns were calculated as the difference between target price and current market price of each security, expressed as a percentage of current market price. Monthly returns, required to determine the covariances, were calculated for each company from the monthly closing prices. The covariance matrix for the 41 stocks was calculated using excel covariance function. The monthly covariance between each pair of securities was converted into annual covariance by multiplying it with 12. The input data of expected returns and covariance matrix were thus made ready for the next step in the analysis.
Portfolio Analysis: The software used is the excel optimizer by Markowitz and Todd (2000) described in the book ‘Mean Variance Analysis and Portfolio Choice’. The software was supplied by Todd on request by the author. The software can handle up to 256 securities.
The software requires as input the expected returns of each security, covariance matrix for the set of securities from which the portfolio is to be formed, lower and upper bounds for the proportion of each security in the portfolio and additional constraints if any.
In the first alternative, the portfolio analysis was done with lower and upper boundary for investment in a single security as zero (zero percent) and one (100 percent) respectively. The additional constraint specified is that the sum of the proportions of all securities has to be one or 100%, the amount available for investment. In the second alternative, the analysis was done with the constraint for individual security holding for mutual funds in India, which is a maximum of 10% of the portfolio in a single security. In this case, the lower and upper bounds are 0 and 0.1. The constraint that the sum of all proportions add to 1 or 100% remains.
RESULTS AND FINDINGS
Corner portfolios describe the efficient frontier. Between any two adjacent corner portfolios, the efficient frontier is a straight line, a weighted average of the two corner portfolios. For alternative 1, the analysis returned 23 corner portfolios. The minimum return portfolio has an expected return of 13.54% and standard deviation of 14.35%. The maximum return portfolio has an expected return of 95.96% and standard deviation of 36.12%.
Investor has to decide the risk level (standard deviation) he wants to bear to select the optimal portfolio from this efficient frontier. This action involves consultation with financial planners. For illustration, if the investor chooses a risk level of 20.27%, the corner portfolio number ‘9’ becomes the optimal portfolio. The expected return of this portfolio is 55.98%. The portfolio is a combination of 9 shares. The proportion or percentage recommended for investment in various securities being:
1. X(2) = 3%
2. X(3) = 13%
3. X(9) = 30%
4 X(14) = 3%
5. X(16) = 35%
6. X(17) = 4%
7. X(34) = 9%
8. X(38) = 2%
9. X(40) = 1%
The total adds up to 100%.
In the second case, the restriction is that upper bound, the proportion invested in any single company’s equity shares, is to be less than 10% of the NAV of the scheme. Accordingly lower bound is specified as zero and upper bound is specified as 0.10. 52 corner portfolios form the efficient frontier in this alternative. The minimum return portfolio has an expected return of 14.02% and standard deviation of 15.59%. The maximum return portfolio has an expected return of 50.64% and standard deviation of 29.35%. It is interesting to compare risk-return characteristics of the maximum return portfolio of alternative 2 with the portfolio selected as an illustration in alternative 1 (55.98% and 20.27%). The expected return is more and standard deviation is lower in the latter case. Thus the constraints imposed through regulation on mutual fund investment are generating an inferior or suboptimal portfolio in this case.
The performance of these two portfolios is compared over one year period from July 05 to June 2006. The mutual fund portfolio (Exp. Ret: 50.64% and Risk: 29.35%) shows a return of 58.4% with 23.13% standard deviation. The other portfolio (Exp. Ret: 55.98% and Risk 20.27%) shows a return of 21.25% with a standard deviation of 21%. As the returns are expected to be more unstable and risk measures are expected to be relatively more stable, the observed performance can be rationalized in such a simple comparison of performance of the two portfolios over one period. Empirical studies to evaluate the superiority of one-year horizon optimal portfolios formed using quantitative methods have to use number of one year periods in the sample.
CONCLUSION AND FUTURE SCOPE FOR RESEARCH
Markowitz’s portfolio analysis can be operationalized and applied to real life portfolio decisions. The 12-month ahead target prices being published for various securities by security analysts can be used as the input for determining expected returns over the next 12 months. The optimal portfolios generated by the portfolio analysis represent the optimal policy for the investor who wants to use the target price estimates rationally.
Acceptance of the methodology for developing and revising portfolios based on target prices provides scope for further research into improving the estimates of the inputs used for portfolio analysis. Also research is to be done to evaluate the performance of the optimal portfolios, in comparison to portfolios formed without using quantitative portfolio analysis models, over a long period of time.
Review of literature reveals that research into the utility of target prices is initiated. Research needs to be extended to find out which target price finding methods are working better. Regarding covariance estimates, Grinold and Kahn (2004) have mentioned that there is possibility of estimation errors in case historical data over a lower number of monthly periods in comparison to number of securities considered for portfolio analysis are used. They suggest structural models. Researchers have to come out with useful models which investors can use on the basis of published data.
Regarding the software for portfolio analysis, the Todd’s program can handle 256 companies. In any particular country, brokers do not normally come out with more than 256 buy recommendations at any point in time. Hence, the software program may not be a limitation. But certainly there will be scope to improve the software, as more and more investors use the methodology, and thereby need efficient and easy to use software with more facilities to come out with various measurements.
For full paper
http://docs.google.com/View?docid=dg3h8m78_2hqm5bkdf
Sources of Data:
Valueline is a monthly bulletin published by Sharekhan (2005) a broking firm in India. The bulletin contains the target price information and the market price on the date of publication for various stocks researched and recommended by the firm. The data from the bulletin of July 2005, which was made available on the website of the firm for public access, is selected for getting the data of expected returns. Target price data was available for 43 companies. Covariance is to be calculated using 25 months closing price data. The monthly closing price data was taken from Prowess, an electronic data base of balance sheet and share price data of Indian companies published by Centre for Monitoring Indian Economy (CMIE, Mumbai). Out of the total 43 companies, for two companies, data was not available for the full 25 months. These two companies were dropped from the set of securities considered for forming the portfolio. Hence, the final list of stocks considered for portfolio analysis contains 41 companies.
Calculation of Input Variables:
The expected returns were calculated as the difference between target price and current market price of each security, expressed as a percentage of current market price. Monthly returns, required to determine the covariances, were calculated for each company from the monthly closing prices. The covariance matrix for the 41 stocks was calculated using excel covariance function. The monthly covariance between each pair of securities was converted into annual covariance by multiplying it with 12. The input data of expected returns and covariance matrix were thus made ready for the next step in the analysis.
Portfolio Analysis: The software used is the excel optimizer by Markowitz and Todd (2000) described in the book ‘Mean Variance Analysis and Portfolio Choice’. The software was supplied by Todd on request by the author. The software can handle up to 256 securities.
The software requires as input the expected returns of each security, covariance matrix for the set of securities from which the portfolio is to be formed, lower and upper bounds for the proportion of each security in the portfolio and additional constraints if any.
In the first alternative, the portfolio analysis was done with lower and upper boundary for investment in a single security as zero (zero percent) and one (100 percent) respectively. The additional constraint specified is that the sum of the proportions of all securities has to be one or 100%, the amount available for investment. In the second alternative, the analysis was done with the constraint for individual security holding for mutual funds in India, which is a maximum of 10% of the portfolio in a single security. In this case, the lower and upper bounds are 0 and 0.1. The constraint that the sum of all proportions add to 1 or 100% remains.
RESULTS AND FINDINGS
Corner portfolios describe the efficient frontier. Between any two adjacent corner portfolios, the efficient frontier is a straight line, a weighted average of the two corner portfolios. For alternative 1, the analysis returned 23 corner portfolios. The minimum return portfolio has an expected return of 13.54% and standard deviation of 14.35%. The maximum return portfolio has an expected return of 95.96% and standard deviation of 36.12%.
Investor has to decide the risk level (standard deviation) he wants to bear to select the optimal portfolio from this efficient frontier. This action involves consultation with financial planners. For illustration, if the investor chooses a risk level of 20.27%, the corner portfolio number ‘9’ becomes the optimal portfolio. The expected return of this portfolio is 55.98%. The portfolio is a combination of 9 shares. The proportion or percentage recommended for investment in various securities being:
1. X(2) = 3%
2. X(3) = 13%
3. X(9) = 30%
4 X(14) = 3%
5. X(16) = 35%
6. X(17) = 4%
7. X(34) = 9%
8. X(38) = 2%
9. X(40) = 1%
The total adds up to 100%.
In the second case, the restriction is that upper bound, the proportion invested in any single company’s equity shares, is to be less than 10% of the NAV of the scheme. Accordingly lower bound is specified as zero and upper bound is specified as 0.10. 52 corner portfolios form the efficient frontier in this alternative. The minimum return portfolio has an expected return of 14.02% and standard deviation of 15.59%. The maximum return portfolio has an expected return of 50.64% and standard deviation of 29.35%. It is interesting to compare risk-return characteristics of the maximum return portfolio of alternative 2 with the portfolio selected as an illustration in alternative 1 (55.98% and 20.27%). The expected return is more and standard deviation is lower in the latter case. Thus the constraints imposed through regulation on mutual fund investment are generating an inferior or suboptimal portfolio in this case.
The performance of these two portfolios is compared over one year period from July 05 to June 2006. The mutual fund portfolio (Exp. Ret: 50.64% and Risk: 29.35%) shows a return of 58.4% with 23.13% standard deviation. The other portfolio (Exp. Ret: 55.98% and Risk 20.27%) shows a return of 21.25% with a standard deviation of 21%. As the returns are expected to be more unstable and risk measures are expected to be relatively more stable, the observed performance can be rationalized in such a simple comparison of performance of the two portfolios over one period. Empirical studies to evaluate the superiority of one-year horizon optimal portfolios formed using quantitative methods have to use number of one year periods in the sample.
CONCLUSION AND FUTURE SCOPE FOR RESEARCH
Markowitz’s portfolio analysis can be operationalized and applied to real life portfolio decisions. The 12-month ahead target prices being published for various securities by security analysts can be used as the input for determining expected returns over the next 12 months. The optimal portfolios generated by the portfolio analysis represent the optimal policy for the investor who wants to use the target price estimates rationally.
Acceptance of the methodology for developing and revising portfolios based on target prices provides scope for further research into improving the estimates of the inputs used for portfolio analysis. Also research is to be done to evaluate the performance of the optimal portfolios, in comparison to portfolios formed without using quantitative portfolio analysis models, over a long period of time.
Review of literature reveals that research into the utility of target prices is initiated. Research needs to be extended to find out which target price finding methods are working better. Regarding covariance estimates, Grinold and Kahn (2004) have mentioned that there is possibility of estimation errors in case historical data over a lower number of monthly periods in comparison to number of securities considered for portfolio analysis are used. They suggest structural models. Researchers have to come out with useful models which investors can use on the basis of published data.
Regarding the software for portfolio analysis, the Todd’s program can handle 256 companies. In any particular country, brokers do not normally come out with more than 256 buy recommendations at any point in time. Hence, the software program may not be a limitation. But certainly there will be scope to improve the software, as more and more investors use the methodology, and thereby need efficient and easy to use software with more facilities to come out with various measurements.
For full paper
http://docs.google.com/View?docid=dg3h8m78_2hqm5bkdf
Piotroski Method
Professor Joseph Piotroski of the Graduate School of Business, Chicago, came out with a much simplified system of fundamental analysis based on the last two years' financial statements that can be used by active investors for picking value stocks with a one- or two-year horizon.
There are nine steps or tests in the model. The stock gets a point for every step/test it passes and a zero for any step it fails. Betting on stocks that score eight or nine points is recommended.
Test 1: Positive net income - Net income, the bottomline after-tax profits, is the simplest measure of profitability. Score a point if the latest year's net income is positive; otherwise, a zero.
Test 2: Positive cash flow - Cash flow is arguably a better profitability measure than net income. Cash flow measures the money that actually moved into or out of a firm's bank accounts. Add one point if the latest year's operating cash flow is positive.
Test 3: Earnings quality - Many experts compare net income to operating cash flow to detect potential accounting manipulations. Cash flow normally exceeds net income because depreciation and other non-cash expenses reduce income, but not cash flow. Award one point if the latest year's operating cash flow exceeds the current year's net income.
Test 4: Decreasing debt - Piotroski rewards companies that are reducing their debt levels. He uses 'financial leverage,' which is total debt divided by total assets, to quantify debt. Award one point if the most recent annual figure is less than the value of the preceding year.
Test 5: Increasing working capital - Working capital, the difference between current assets and current liabilities, measures the cash available to run the business.
Piotroski prefers stocks with increasing working capital. Current ratio, which is current assets divided by current liabilities, is the usual metric for measuring working capital. Award one point if the most recent annual figure exceeds the preceding year's number.
Test 6: Increase in asset turnover - Award one point if the most recent annual asset turnover (sales revenue divided by total assets) exceeds the year-ago figure.
Test 7: Growing profitability - Return on assets measures overall profitability by comparing net income to total assets (net income divided by total assets). Award one point if the most recent annual ROA exceeds the year-ago figure.
Test 8: Issuing stock - Piotroski prefers companies that did not issue more stock to raise capital or to fund acquisitions. Award one point if the most recent number of total shares outstanding is equal to, or less than, the year-ago figure.
Test 9: Competitive position - Increasing competition often forces companies to cut prices, and hence profit margins, to maintain sales. Conversely, rising profit margins signal an improving competitive position. Award one point if this year's gross profit margin exceeds the year-ago number.
Piotroski recommended the use of this method on the lowest quintile of shares on New York Stock Exchange
Piotroski's scoring system is easy to use. It was found to be effective over the period from 1976 to 1996 by Piotroski himself.
There are nine steps or tests in the model. The stock gets a point for every step/test it passes and a zero for any step it fails. Betting on stocks that score eight or nine points is recommended.
Test 1: Positive net income - Net income, the bottomline after-tax profits, is the simplest measure of profitability. Score a point if the latest year's net income is positive; otherwise, a zero.
Test 2: Positive cash flow - Cash flow is arguably a better profitability measure than net income. Cash flow measures the money that actually moved into or out of a firm's bank accounts. Add one point if the latest year's operating cash flow is positive.
Test 3: Earnings quality - Many experts compare net income to operating cash flow to detect potential accounting manipulations. Cash flow normally exceeds net income because depreciation and other non-cash expenses reduce income, but not cash flow. Award one point if the latest year's operating cash flow exceeds the current year's net income.
Test 4: Decreasing debt - Piotroski rewards companies that are reducing their debt levels. He uses 'financial leverage,' which is total debt divided by total assets, to quantify debt. Award one point if the most recent annual figure is less than the value of the preceding year.
Test 5: Increasing working capital - Working capital, the difference between current assets and current liabilities, measures the cash available to run the business.
Piotroski prefers stocks with increasing working capital. Current ratio, which is current assets divided by current liabilities, is the usual metric for measuring working capital. Award one point if the most recent annual figure exceeds the preceding year's number.
Test 6: Increase in asset turnover - Award one point if the most recent annual asset turnover (sales revenue divided by total assets) exceeds the year-ago figure.
Test 7: Growing profitability - Return on assets measures overall profitability by comparing net income to total assets (net income divided by total assets). Award one point if the most recent annual ROA exceeds the year-ago figure.
Test 8: Issuing stock - Piotroski prefers companies that did not issue more stock to raise capital or to fund acquisitions. Award one point if the most recent number of total shares outstanding is equal to, or less than, the year-ago figure.
Test 9: Competitive position - Increasing competition often forces companies to cut prices, and hence profit margins, to maintain sales. Conversely, rising profit margins signal an improving competitive position. Award one point if this year's gross profit margin exceeds the year-ago number.
Piotroski recommended the use of this method on the lowest quintile of shares on New York Stock Exchange
Piotroski's scoring system is easy to use. It was found to be effective over the period from 1976 to 1996 by Piotroski himself.
Graham-Rao Method
The model by Benjamin Graham, who is credited with systematising fundamental analysis, recommended for use by conservative investors.
Accordng to Graham's stock-selection criteria, the company must have an adequate size (Rs 100 crore sales may be taken as adequate size for Indian companies) and a strong financial condition.
To satisfy this criterion, the current assets should be at least twice that of current liabilities and the total debt-equity ratio should not be greater than 1:1.
The company should have paid dividends and earned profits for the last 10 years.
There should be a growth in earnings per share of 10 per cent per annum over the last seven years.
The current price should not exceed 20 times the average EPS in the last seven years for companies with past seven-year growth higher than 20 per cent. For companies with past growth rates between 10 and 20 percent per annum, the multiplier has to be the growth rate itself.
The current price should also not be more than 1.5 times the book value last reported.
These prescriptions by Graham require 10-year data to pick stocks. But the method is unambiguous and uses a limited number of ratios.
Investors may complain about the 10-year data requirement; but they have to keep in mind that their hard-earned money has to be protected by committing it to companies with a good past record.
Accordng to Graham's stock-selection criteria, the company must have an adequate size (Rs 100 crore sales may be taken as adequate size for Indian companies) and a strong financial condition.
To satisfy this criterion, the current assets should be at least twice that of current liabilities and the total debt-equity ratio should not be greater than 1:1.
The company should have paid dividends and earned profits for the last 10 years.
There should be a growth in earnings per share of 10 per cent per annum over the last seven years.
The current price should not exceed 20 times the average EPS in the last seven years for companies with past seven-year growth higher than 20 per cent. For companies with past growth rates between 10 and 20 percent per annum, the multiplier has to be the growth rate itself.
The current price should also not be more than 1.5 times the book value last reported.
These prescriptions by Graham require 10-year data to pick stocks. But the method is unambiguous and uses a limited number of ratios.
Investors may complain about the 10-year data requirement; but they have to keep in mind that their hard-earned money has to be protected by committing it to companies with a good past record.
Subscribe to:
Posts (Atom)